Content uploaded by Cheryl L Carmichael
All content in this area was uploaded by Cheryl L Carmichael on Mar 25, 2015
Content may be subject to copyright.
With a Little Help From My Friends:
Long-term Self-perceived Health, Neuroendocrine,
and Well-being Correlates of Early Adult Social Activity
Cheryl L. Carmichael
Submitted in Partial Fulfillment
Requirements for the Degree
Doctor of Philosophy
Professor Harry Reis
Department of Clinical and Social Sciences in Psychology
Arts, Sciences and Engineering
School of Arts and Sciences
University of Rochester
Rochester, New York
For my late parents, Robert and Francine Carmichael. Their unconditional
love, support, and belief in me gives me the purpose to pursue my path, and the
courage to stay the course even when the path grows dark, and the way is dim.
Though their time was cut short, the love, kindness, and generosity they showed to all
carries on in the lives they touched. They set an example to live by, and the world is a
better place for them having been in it.
Cheryl Carmichael was born in Schenectady, New York on March 15, 1978.
She matriculated at Northeastern University in 1996 and graduated in 2001 with a
Bachelor of Science degree in Psychology. She came to the University of Rochester
in the Fall of 2001 and began graduate studies in Social Psychology. She pursued her
research in close relationships under the direction of Professor Harry Reis, and
received a Master of Arts degree from the University of Rochester in 2005. She was
awarded a National Science Foundation Research Experience for Undergraduates
supplement in 2000 to study attentional biases to emotion with Judith Hall. She was
an instructor at St. John Fisher College in Rochester, NY from 2005 to 2007, and was
the 2006 recipient of the Vincent and Helen Nowlis Award for Excellence in
Teaching. In 2007 she and Dr. Reis were awarded a Rochester Center for Mind-Body
Research pilot grant to conduct this research on the long-term correlates of early adult
social activity, which marks the culmination of her academic training. She was a
visiting scholar at the Columbia University Graduate School of Business from 2009
to 2010, and currently resides in Brooklyn, New York where she is an instructor at
Fordham University and City University of New York, Brooklyn College.
List of Publications:
Lane, R.D., Carmichael, C.L., & Reis, H.T. (in press). Differentiation in the
momentary rating of somatic symptoms covaries with trait emotional awareness
in patients at risk for sudden cardiac death. Psychosomatic Medicine.
Tomlinson, J.M., Carmichael, C.L., Reis, H.T., & Aron, A. (2010). Affective
forecasting and individual differences: Accuracy for relational events and anxious
attachment. Emotion, 10, 447-453.
Reis, H.T., Smith, S.M., Carmichael, C.L., Caprariello, P.A., Tsai, F.F., Rodrigues,
A., & Maniaci, M.R. (2010). Are you happy for me? How sharing positive events
with other provides personal and interpersonal benefits. Journal of Personality
and Social Psychology, 99, 311-329.
Carmichael, C.L., Tsai, F.F., Smith, S.M., Caprariello, P.A., & Reis, H.T. (2007).
The self and intimate relationships. In C. Sedikides & S. Spencer (Eds.) The Self.
(pp. 285 – 309). London: Psychology Press.
Carmichael, C.L., & Reis, H.T. (2005). Attachment, sleep quality, and depressed
affect. Health Psychology, 24, 526-531.
I am so very grateful to everyone who helped to make this dissertation
possible; the list is long, and for that I consider myself lucky! First and foremost, I
wish to extend heartfelt thanks to my advisor, Harry Reis. It has been my privilege
and pleasure to work with someone who is so instinctually attuned to the field, and
has been so unreservedly willing to share his vast experience and soaring intellect
with me. I am indebted to him for his generosity. He has indelibly influenced my
development as researcher, and his devotion to his work continues to inspire and
impress me. Harry has been an amazing mentor and friend, and I am grateful for the
depths of encouragement, support, patience, and care he has shown me.
This project is better off for the contributions of my committee members, Paul
Duberstein and Miron Zuckerman, thanks to both for investing in my intellectual
growth. Additional thanks to Miron for his unassuming approach to mentoring, and
for the warmth and understanding that he showed me.
Several people provided administrative and technical support that was vital to
this project. My appreciation to Jan Moynihan for her advice on cortisol sampling; to
April Engram for processing participant payments, and collecting saliva samples from
my mailbox in my absence; to James Walton for finding the time not only to assay
my saliva samples, but also allowing me to participate in the process and patiently
teaching me about immunoassays; to Nathan Franus for being a kind guide when I
was new to the RCMBR; and to Patrick Davies and Karin Gasaway for sharing
freezer space. I wish to extend thanks to Ilene LaMalfa, Karen Medalle, Audrey
Nolan, and Loretta Pratt who so generously volunteered their time to pilot test the
surveys used in this research. I am especially indebted to Maryann Gilbert for her
tireless assistance and support, and for easing what would have been an otherwise
harrowing process given my circumstances.
This work could not have been completed without the generous support of a
Rochester Center for Mind-Body Research pilot grant, funded by National Institute
on Aging grants R24AG031089 and R21AG023956, or the research participants who
gave of their time and effort with more enthusiasm than I could have hoped for.
My family has been a source of great strength and encouragement. I am
thankful for the support of my brother, John, and for my Aunt Janet and Uncle Gary’s
unwavering faith in my ability to succeed. I am also grateful to Dana Carney, my first
mentor, and now lifelong friend. I became a scientist because of her. Thanks also to
my fellow graduate students Denise Gettman, Shannon Smith, Amy Rodrigues, Feng-
Fang Tsai, Tony Monacelli, Marcia Winter, Ryan O’Loughlin, Peter Caprariello, and
Mike Maniaci. They enriched my life in Rochester and my experience in graduate
school; I am fortunate to count them among my friends.
Finally, I am eternally grateful to my patient, caring, and giving partner,
Marty. His whole-hearted support of every kind was immeasurable, and I could not
have done without it. He cultivated joy, laughter, and fun in our lives, even in the
bleakest of times. He unfailingly believed in my capabilities, and stood by my side
through every obstacle I encountered, absorbing the impact to shelter me from the
storm and shepherd me forward. I get by with a little help from him.
This research investigated longitudinal associations between social activity
during college (a phase of early adulthood singled out among developmental theorists
as crucial to relationship development), and social, emotional, and health outcomes in
midlife. Men and women age 48-50 (n = 133) were recruited from a pool of
University of Rochester graduates who completed two-weeks of event-contingent
social interactions records during college in the late 1970s. College interaction diaries
provided measures of structural social integration (i.e., interaction quantity), and
functional perceived support (i.e., interaction quality), which were simultaneously
tested as predictors of midlife health and well-being. Recruited participants reported
on current social relationship quality, emotional adjustment, and physical health, and
provided a series of saliva samples to capture diurnal cortisol rhythm. Midlife social
relationship quality, emotional adjustment, and self-reported health were positively
interrelated. Longitudinally, college social interaction quantity predicted enhanced
social (friendship quality, intimacy with romantic partner, social network size), and
emotional (diminished negative affectivity, enhanced positive affectivity and self-
actualization) outcomes 30 years later, in midlife. Perceived social relationship
quality in college was longitudinally associated with enhanced social (friendship
quality, romantic intimacy), and health (better cardiovascular health) outcomes in
midlife. The discussion focuses on possible mechanisms responsible for longitudinal
associations that span 30 years, the relative impact of structural versus functional
support, and the general invariance of these effects across sex.
Table of Contents
Curriculum Vitae iii
Table of Contents viii
List of Tables xii
List of Figures xv
Chapter 1: Introduction 2
Relationships and Health: The Good, The Bad, and The Absence 5
The absence: The harms of social isolation 6
Social engagement: Structure or function? 7
The bad: The negative consequences of aversive relationships 11
The good: The benefits of constructive relationships. 12
A Life Span Approach 19
Aging and health 19
Early adulthood to midlife 21
Social relationships as the mechanism 23
The Current Research 26
The necessity of longitudinal designs 27
The existing sample 29
Experience sampling methods 30
The current assessment 32
Social well-being hypothesis 35
Emotional adjustment hypothesis 35
Health hypothesis 35
Chapter 2: Method 36
College Measures 38
Midlife Measures 38
Emotional adjustment measures 39
Social well-being measures 40
Self perceived health measures 42
Other measures 43
Neuroendocrine function 44
Sample storage and shipping 46
Chapter 3: Results 47
Participant attrition and comparability 47
Data Reduction 48
Social well-being measures 49
Emotional adjustment measures 50
Physical health measures 50
Neuroendocrine function 51
Sex, Original Sample, and Freshman Versus Senior Status 53
Concurrent Analyses 55
Sex differences in concurrent correlations 57
Longitudinal Prediction of Midlife Outcomes 58
Analysis strategy 58
Social life 58
Emotional adjustment 62
Physical health 64
Neuroendocrine function 66
Summary of longitudinal findings 67
Controlling midlife social relationship quality 67
Controlling personality 68
Chapter 4: Discussion 71
Summary of Concurrent Findings 71
Summary of Longitudinal Findings 74
Unsupported Longitudinal Hypotheses 77
Romantic relationship quality 77
Physical health 80
Measurement concerns 81
Health status 84
His, Hers, or Theirs: Sex Differences 86
Assessing Support: Form Versus Function 89
Conclusions and Future Directions 99
Appendix A 161
List of Tables
Table Title Page
1 Summary of function and measurement of three
physiological systems 123
2 Summary of previous samples and recruitment for
current study 124
3 Comparison of recruited versus unrecurited participants
on college social activity variables 125
4 Factor loadings for individual social relationship quality
scales on three principle components with oblimin rotation 126
5 Factor loadings for individual emotional adjustment scales
on three principle components with oblimin rotation 127
6 Factor loadings for individual physical health scales on
three principle components with oblimin rotation 128
7 Effects of original sample and year in college (freshman
vs. senior) on midlife social well-being 129
8 Effects of original sample and year in college (freshman
vs. senior) on midlife emotional adjustment 130
9 Effects of original sample and year in college (freshman
vs. senior) on midlife physical health 131
10 Effects of original sample and year in college (freshman
vs. senior) on neuroendocrine function 132
11 Concurrent correlations among midlife social, emotional,
and physical health composite variables 133
12 Concurrent correlations among midlife social well being,
emotional adjustment, physical health and neuroendocrine
composites and individual measures of social well being and
emotional adjustment 134
Table Title Page
13 Concurrent correlations among midlife social well being,
emotional adjustment, physical health and neuroendocrine
composites and individual measures of physical health
and neuroendocrine function 135
14 Concurrent correlations among midlife social well being,
emotional adjustment, physical health, and neuroendocrine
function composites for men (below the diagonal) and
women (above the diagonal) 136
15 Prediction of midlife romantic relationship quality from
college social activity quantity and quality 137
16 Prediction of midlife friendship quality from college social
activity quantity and quality 138
17 Prediction of midlife social network from college social
activity quantity and quality 139
18 Prediction of midlife negative affectivity from college
social activity quantity and quality 140
19 Prediction of midlife positive affectivity from college
social activity quantity and quality 141
20 Prediction of midlife self-actualization from college social
activity quantity and quality 142
21 Prediction of midlife daily symptomatology from college
social activity quantity and quality 143
22 Prediction of midlife cardiac health I from college social
activity quantity and quality 144
23 Prediction of midlife cardiac health II from college social
activity quantity and quality 145
24 Prediction of midlife neuroendocrine function from
college social activity quantity and quality 146
Table Title Page
25 Summary of effects of overall early adult social activity
on midlife composite social well being, emotional
adjustment, and physical health variables 147
26 Summary of effects of overall early adult social activity on
midlife composite social well being, emotional adjustment,
physical health, and neuroendocrine function variables
controlling for midlife social well being variables 148
27 Summary of effects of overall early adult social activity
on midlife composite social well being, emotional
adjustment, physical health, and neuroendocrine function
controlling for midlife emotional stability/neuroticism and
List of Figures
Figure Title Page
1 Hypotheses 150
2 Original Sample X Participation Status interaction on
number of opposite sex interactions per day 151
3 Original Sample X Sex interaction on midlife daily
4 Original Sample X Sex interaction on cardiac health II 153
5 College Year X Sex interaction on positive affectivity 154
6 Same-sex Quantity X Sex interaction on perceived
partner responsiveness 155
7 Overall Quantity X Sex interaction on perceived social
support from friends 156
8 Same-sex Quality X Sex interaction on social network size 157
9 Overall Quantity X Sex interaction on personal growth 158
10 Overall Quality X Sex interaction on cardiac diagnosis history 159
11 Overall Quality X Sex interaction on cardiac health II
The author collected all of the data in the current assessment.
Cortisol assays were performed by James Walton, and assisted by the author, at the
Rochester Center for Mind-Body Research, Strong Memorial Hospital, University of
Rochester, Rochester, NY.
College social activity data were collected by Ladd Wheeler (samples I, II, & III),
Harry Reis (samples II & III), and John Nezlek (samples I & II).
Chapter 1: Introduction
There is no hope of joy except in human relations. –Antoine de Sainte-Exupéry
The list of maxims that one could collect enumerating the benefits of good
relationships would undoubtedly be lengthy. These incisive, yet simple observations
conceived across the years by philosophers, writers, musicians, and lay people alike,
are substantiated by the scientific literature; socially integrated individuals tend to
live longer (e.g., Holt-Lundstad, Smith, & Layton, 2010; House, Landis, &
Umberson, 1988) and tend to be better psychologically adjusted (e.g., Argyle, 1999;
Umberson, Chen, House, Hopkins, & Slaten, 1996).
Showing that characteristics of social life are systematically associated with
health and longevity gives rise to several important questions about how and why
such benefits might materialize. First, are there causal links between social life and
health outcomes? Although experimental manipulation is the gold standard for
establishing causality, manipulating certain variables (such as relationship quality, for
instance) is more than an issue of adequate research design; it is also a question of
ethics and feasibility. Our ability to make causal inferences about the relationship
between social life and health hinges on well-designed prospective longitudinal
investigations of social life.
Second, if relationships are causally linked to health, what are the mechanisms
by which such an effect may emerge? It remains to be seen whether health varies as a
function of immediate social circumstance, or reflects something more long standing,
such as an accumulation of benefits acquired over the life course. For example,
emergent research has linked positive emotional expression in early adulthood with
later life health and well being (Danner, Snowden, & Friesen, 2001; Harker &
Keltner, 2001), and positive social relationships are the hypothesized mediating
mechanism. However, little evidence exists to bear out this claim. Perhaps an
existence imbued with warm and supportive relationships provides opportunities for
positive emotional expression, promotes psychological well-being and adaptive
coping, and thus enhances health and longevity. These possibilities remain yet
untested because few studies have examined these associations longitudinally, and
even fewer have contemporaneously examined social life during early adulthood, the
period most critical to intimate relationship development (Arnstein, 1984; Erikson,
Despite some provocative and suggestive findings, our understanding of the
links between social connection and health remains muddied by several limiting
factors. First, this type of research has been largely conducted among a limited birth
cohort. It remains to be seen whether they will generalize to those who were born a a
different time, and lived through a different set of historical events and societal
conditions. In addition, for individual studies, effect sizes are rather small, and the
results of various studies do not always converge on the same conclusions. For
example, the literature diverges with regard to whether men or women benefit equally
from social integration (Keicolt-Glaser & Newton, 2001). Some studies have found
the benefits of social relationships to be relatively greater for men (Diener, Suh,
Lucas, & Smith, 1999; Keicolt-Glaser & Newton, 2001), whereas others have
suggested quite the reverse (Glenn, 1975; Wood, Rhodes, & Whelan, 1989).
The purpose of the current investigation is to adopt a lifespan approach, and
longitudinally examine the midlife health and well-being correlates of individuals’
social activity in college (referred to as early adulthood for convenience). Several
theories of lifespan development identify early adulthood as the critical stage for the
flourishing of intimacy and close social relationships (Erickson, 1959; Levinson,
1978; Neugarten, Moore, & Lowe, 1965; White, 1966). Moreover, Hartup and
Stevens’ (1997) argue that throughout life, friends play a vital role in adaptation by
fostering a sense of well being, and facilitating adjustment to life’s transitions.
Stability in social activity from college to age 30 (Reis, Lin, Bennett, & Nezlek,
1993) may connote stability throughout adult life. If key skills that are cultivated
during college can be carried through early adulthood, there is good reason to believe
these skills will continue to be employed throughout the remainder of adult life. Thus,
those with an enriched social network in college may be armed with tools that help
them reap lifelong benefits of good social ties. Those lifelong benefits may include
the provision of social support, an outlet for the experience and expression positive
affect, encouragement of healthy behaviors, and attenuation of the negative effects of
stress, all of which may promote the long-term welfare of recipients (Berkman, Glass,
Brissette, & Seeman, 2000). To investigate whether early adulthood social activity is
related to midlife health and well being, adults (approximately 50 years of age) who
participated in detailed social interaction diary research during college will be
recruited for an assessment of current self-perceived health, neuroendocrine function,
and social and emotional well being.
In the following pages I will first provide a detailed overview of the
connection between social life and health and well-being outcomes. I will then
explain why midlife presents an opportune time for examining health and well being.
I will argue for the utility of a lifespan approach in this domain, and provide a
theoretical rationale for linking social activity in early adulthood to outcomes in
midlife. Next, I will describe the characteristics of the target sample and explain the
unique features that are provided by using a longitudinal design with this particular
group of individuals. Finally, I will offer hypotheses about the relationship between
social engagement in early adulthood and health and well being in midlife.
Relationships and Health: The Good, the Bad, and the Absence
Berkman et al. (2000) present an overarching framework for linking social life
to health whereby large-scale factors such as culture and social network size may
facilitate or hinder the availability of intermediate psychosocial mechanisms such as
social engagement, person-to-person contact, and social support. Psychosocial
mechanisms, in turn, are hypothesized to have an effect on health through a number
of specific physiological (e.g., cardiovascular reactivity) and psychological (e.g.,
depression) pathways. Their model suggests that the presence of relationships is
necessary, but not sufficient, for the attainment of benefits. In other words,
relationships provide the means for subsequent events that have more proximal
implications for health. Consistent with this idea, in the following sections I will
selectively review research that supports three broad conclusions with regard to
health and well being outcomes: 1) a lack (or perceived lack) of relationships may be
harmful; 2) caustic relationships seem to produce negative consequences; and 3)
positive social relations appear to have beneficial effects.
The broader term health has been employed to refer to a great many specific
and interrelated outcomes, both subjective and objective. The scope of the current
project encompasses self-perceived healthfulness, neuroendocrine functioning, and
psychological well-being. However, when reviewing the empirical evidence in
support of the three claims above, I will report findings beyond the scope of the
current investigation. In addition to reviewing psychological health and well-being
outcomes, I will review physiological outcomes for the cardiovascular system, the
immune system, and the neuroendocrine system, which includes the autonomic
nervous system and the hypothalamic-pituitary-adrenal (HPA) axis (see Table 1 for a
summary of major functions of these three systems, and the most common measures
associated with each). Although the current project will examine neuroendocrine
function as the sole measure of objective health status, I will maintain a broader focus
in my review because these interrelated systems are critical to the physiological
processes of homeostasis (maintaining consistency), allostasis (maintaining
consistency through change), and thus long-term health (McEwen, 1998), a point that
will be elaborated upon in the next section.
The absence: The harms of social isolation. A seminal paper by Berkman
and Syme (1979) showed that the presence of each of four types of social ties (a
spouse, extended family and friends, a church, and other formal and casual groups)
was associated with reduced mortality risk up to nine years later among residents of
Almeida County, California. These findings remained significant when controlling
various self-reported mortality-relevant variables such as physical health, smoking,
alcohol consumption, obesity, socioeconomic status, use of preventative health
services, and life satisfaction. Although the use of self-reports arguably undermines
the credibility of these findings, the pattern has been replicated in subsequent
prospective studies with culturally diverse samples across the United States and
Europe. Social integration is positively and consistently associated with reduced
mortality and morbidity rates (Berkman, 1995; House, Landis, & Umberson, 1988).
An examination of the research on social integration and health reveals
variations in the measurement of social integration. Some researchers ask about the
number of relationships one has, others ask about the frequency of social interaction,
and yet others ask about subjective feelings of closeness and connection with others.
Thus, it is important to be clear about what is meant by social engagement, and the
outcomes predicted by each conceptualization.
Social engagement: Structure or function? Among the ways to quantify
social connectedness, two seem to predominate. The structural approach examines
social integration in a relatively objective or quantitative manner; that is, by looking
at the presence versus absence of specific social relationships (e.g., marital partner,
extended family, neighbors, etc.), the number of social ties one has, and the frequency
with which one interacts with those social contacts. As described above, research
stemming from this paradigm demonstrates that structural social integration is
associated with physical health gains (e.g., Berkman, 1995; Berkman & Syme, 1979;
House, Landis, & Umberson, 1988). For example, individuals with fewer types of
social connections were found to be more susceptible to the common cold than their
socially integrated counterparts. In fact, those with the fewest kinds of social ties
were 4.2 times as likely (adjusted relative risk ratio) to contract a cold as those with
the greatest diversity of social ties. Moreover, controlling health behaviors (e.g.,
smoking and exercise) did not eliminate the effect of social integration on cold virus
susceptibility (Cohen, Doyle, Skoner, Rabin, & Gwaltney, 1997). Even temporary
bouts of isolation may have negative health consequences. A 4-7 day separation from
one’s romantic partner resulted in heightened salivary cortisol levels among those
high in attachment anxiety (Diamond, Hicks & Otter-Henderson, 2008). High levels
of salivary cortisol reflect HPA axis dysregulation, which may have negative
implications for immune system function and thus the body’s ability to fight disease
The functional approach, on the other hand, examines social life from a
relatively subjective standpoint; that is, by looking at the perceived availability of
social support and feelings of loneliness and belonging, apart from the number of
social relationships an individual may have. Research stemming from this paradigm is
consistent with findings from the structural approach; when the functions fulfilled by
social relationship are unsated, health and well being suffer. For example, Cohen,
Mermelstein, Kamarck, and Hoberman (1985) reviewed studies that showed a
perceived lack of available social support predicted poorer mental and physical health
outcomes. This was especially true for appraisal support (believing one has someone
to discuss problems with), esteem support (believing one has someone to make a
positive social comparison with), and belonging support (believing one has someone
to spend time with), but less clear for tangible support (believing one has someone
who will provide material aid). More specifically, in a study of metastatic breast
cancer patients, women who felt they had few relatives or friends with whom they
could discuss private matters showed elevations in mean salivary cortisol levels
(Turner-Cobb, Sephton, Koopman, Blake-Mortimer, & Spiegel, 2000).
Loneliness, another functional measure of integration, has garnered a great
deal of recent attention as a predictor of health-relevant outcomes. Feelings of
loneliness may be partly attributed to a discrepancy between an individual’s desired
and actual social circumstances (Peplau & Perlman, 2000). Lonely people (relative to
non-lonely people) exhibit a variety of physiological differences that may contribute
to impoverished health. Lonely people are less likely to enjoy the restorative benefits
of sleep (Cacioppo et al., 2000), and display a lower basal heart rate, which may
reflect emotional withdrawal (Cacioppo et. al., 2000). Moreover, they respond poorly
to stressors, exhibiting increased total peripheral resistance (TPR) and decreased
cardiac output (CO; Cacioppo et al., 2002), a pattern associated with inhibited cardiac
recovery from stress (Dienstbier, 1989; Mendes, Reis, Seery, & Blascovich, 2000).
Lastly, the chronically lonely display increased levels of salivary cortisol (indicating a
higher level of stress), and show maladaptive HPA axis activity whereby cortisol
levels are elevated in the evening (when cortisol is typically on the decline and lowest
for most people; Cacioppo et al., 2000). In short, across the body’s systems, the
psychophysiological costs borne by lonely people may contribute to poorer health.
Whereas structural measures of support are relatively objective, functional
measures of support capture subjective interpretations of social life quality, and tend
to vary as a function of more enduring aspects of a person, such as their personality,
self-esteem, or attachment orientation. A body of research shows that secure,
emotionally stable, high self-esteem individuals are more likely to feel supported; to
believe that the close others in their lives are reliabile, trustworthy, and will be there
when needed; and perceive the objectively same support-relevent behaviors as more
comforting and supportive than their less secure counterparts (Murray, Rose,
Bellavia, Holmes, & Kusche, 2002; Shaver & Mikulincer, 2002). Nevertheless,
studies of structural and functional measures of social integration converge on similar
conclusions; social isolation, real or imagined, is associated with impoverished health
and well being. Although structural social isolation may give rise to feelings of
loneliness (Wenger & Burholt, 2004) and a diminished perception of functional
support, the relationship is not perfect. Relationship partners can be responsible for
feelings of loneliness or discontent when conflict arises or rejection occurs, and
spending time alone may have restorative benefits (Kiecolt-Glaser, Glaser, Cacioppo,
& Malarkey, 1998). This point underscores two important considerations. First, it
highlights a key issue that remains to be addressed. Do structural and functional
social networks produce similar effects when measured simultaneously? The current
research will attempt to shed light on this question by examining both structural
measure of integration, and functional measures of perceived relationship quality.
College social activity will be assessed in terms of both quantity of social engagement
and quality of social interactions. Second, it calls attention to the potentially harmful
aspects of close relationships.
The bad: The negative consequences of aversive relationships. Although
the absence (or perceived absence) of social connection may be detrimental to health,
simply having relationships is not uniformly beneficial. Hartup and Stevens (1997)
note that “…good outcomes are most likely when one has friends, one’s friends are
well socialized, and when one’s relationships with these individuals are supportive
and intimate.” (p. 365). Negative social relationships can be just as (if not more)
harmful as being or feeling alone.
Negative marital function is associated with poor outcomes for spouses.
Destructive marriages indirectly impact individuals’ health through depression and
poor health habits (Keicolt-Glaser & Newton, 2001). In addition, dissatisfaction and
doubt surrounding the relationship with one’s spouse is associated with sleep
disruptions (Carmichael & Reis, 2005), and poorer subjective health (Levenson,
Carstensen, & Gottman, 1993). Furthermore, marital conflict appears to take a direct
toll on the cardiovascular, neuroendocrine, and immune systems. (Kiecolt-Glaser &
Newton, 2001). During conflict discussion dissatisfied wives evinced higher systolic
blood pressure and heart rate than satisfied wives (Carels, Sherwood, & Blumenthal,
1998). This sort of cardiovascular reactivity to stress may contribute to the
development of cardiovascular disorders (Manuck, 1994), and coronary heart disease
(Orth-Gomer, Rosengren, & Wilhelmsen, 1993). Moreover, displays of hostility or
other negative behaviors (e.g., criticism, denying responsibility, making excuses,
negative mind reading) during conflict are associated with poor physiological
responses for both partners. For instance, negative behaviors during conflict were
associated with the production and secretion of catecholamines, which are HPA axis
hormones generated in response to stress (Malarkey, Kiecolt-Glaser, Pearl, & Glaser,
1994). Those same types of negative behaviors have also been linked to down-
regulation of the immune system (Kiecolt-Glaser, Malarkey, Chee, Newton,
Cacioppo, Mao, & Glaser, 1993). Neuroendocrine and immune system dysfunction
may have negative repercussions for health and well being, especially if conflict
Notwithstanding the problems that negative relationships may create, poor
quality marital relations may hinder disease recovery. Myocardial infarction (MI)
patients who perceived their marital partners to be relatively unsupportive were more
likely to report post-MI chest pain, and were also more likely to have been
rehospitalized up to one year after the initial heart attack (Helgeson, 1991).
Considerable evidence indicates that negative social relationships may
interfere with adaptive psychophysiological responding, and thus good health.
However, given the health gains associated with structural integration (e.g., Holt-
Lundstad, Smith, & Layton, 2010; House, Landis, & Umberson, 1988), positive
social interactions evidently may be beneficial to health and well being.
The good: The benefits of constructive relationships. Whereas negative
social relationships and feelings of social disconnectedness appear to be harmful, a
variety of benefits may be reaped from caring and supportive close relationship
partners who are able provide intimate social contact in the face of life’s ups and
downs. Positive social relationships may serve as sources of self-worth and well
being throughout all stages of life (Hartup & Stevens, 1997), a known factor in
cardiovascular and neuroendocrine health (Ryff et al., 2006).
It is important to note that apparent benefits of good relationships are believed
to be more than just an absence of harm associated with bad relationships. Research
by Uchino and colleagues suggests that a relationship partner can be supportive
(mostly positive), aversive (mostly negative), or ambivalent (a combination of the
two; Uchino, Holt-Lunstad, Uno, & Flinders, 2001). In other words, rather than a
model consisting of a single dimension with “good” and “bad” as opposite ends of the
same continuum, a bivariate model is more appropriate. Bivariate models are
comprised of two independent dimensions, in this case, one dimension representing
relational positivity (running from low to high), the other representing relational
negativity (also running from low to high). Where a partner falls on each dimension is
unrelated to his or her position on the other dimension. Research in a variety of
domains supports bivariate models, despite frequent inverse associations between the
two dimensions. This has been demonstrated for attitudes and affective experiences
(Cacioppo & Berntson, 1994; Larsen, McGraw, & Cacioppo, 2001), in the motivation
literature (Gable, Reis, & Elliot, 2003), and more recently the domain of
psychological well being (Ryff et al., 2006). Although Uchino et al.’s (2001) research
has largely focused on outcomes associated with ambivalent network ties (those who
are high in both positivity and negativity), their research demonstrates that positive
and negative aspects of relationships are independent, and benefits of good
relationships may not simply reduce to the absence of harm from bad relationships.
The literature distinguishes two models for explaining the ways in which
social support networks consisting of friends and loved ones may be beneficial to
health: the stress-buffering model, and the main effect model. The stress-buffering
model states that social support networks produce health benefits by reducing the
pathogenic effects of stress (Cassel, 1976; Cobb, 1976). As I will discuss shortly,
stress is a leading contributor to health throughout the life course, and the negative
effects of stress may accumulate over time creating a long-term adverse impact on
biological functioning (McEwen, 1998). According to this model, in the absence of
stress, support is irrelevant. However, when stress is present, relationship partners
may act as a protective buffer, promoting health by alleviating the mental and
physical toll of stress on the individual. This is consistent with Hartup and Stevens’s
(1997) argument that close others help individuals adjust to changing circumstances
and adapt to life’s stressors from childhood, through adolescence, and even into
Supportive social interactions have been shown to reduce cardiovascular
reactivity to stressors in the laboratory (Uchino, Cacioppo, & Kiecolt-Glaser, 1996).
When participants were given a stress induction, those who perceived high levels of
support showed the typical pattern of cardiovascular reactivity, a relative decrease in
reactivity among older versus younger participants. However, among those who
perceive low levels of social support the reverse was true; older participants showed
relatively increased cardiovascular reactivity in response to stress (Uchino, Kiecolt-
Glaser, & Cacioppo, 1992). Likewise, Gump, Polk, Kamarck, and Shiffman (2001)
examined ambulatory blood pressure in a naturalistic setting. They found that
interactions with a committed romantic partner, compared to interactions with any
other persons and non-interaction, were associated with reduced ambulatory blood
pressure. In short, supportive relationships can mitigate the harmful cardiovascular
effects of stress and promote healthy cardiovascular functioning.
Partners may also help buffer the damaging effects of stress on
neuroendocrine function. Stress activates the HPA axis, producing elevated levels of
cortisol, which can be harmful to the brain and other organs with prolonged or
persistent exposure (Sapolsky, 1990; 1996). Whereas a lack of social support is
associated with neuroendocrine dysregulation and elevations in cortisol levels
(Cacioppo et al., 2002; Turner-Cobb et. al., 2000), positive social engagement has
been shown to engender adaptive HPA axis responses to stress. During a conflict
discussion task, husbands’ positive behavior was associated with sharper declines in
two stress-related hormones (adrenocorticotropic hormone and cortisol) among wives
(Robles, Shaffer, Malarkey, & Keicolt-Glaser, 2006). This efficient recovery pattern
suggests that constructive social interaction promotes adaptive physiological
responding, which could potentially protect against the deleterious effects of stress.
In contrast to the evidence that negative social relationships appear to impede
disease recovery (Helgeson, 1991), positive social contact has been shown to
facilitate disease recovery. For example, coronary artery surgery patients who
received support that demonstrated caring and respect exhibited lower levels of
functional disruption and higher levels of emotional well-being up to one year after
surgery (King, Reis, Porter, & Norsen, 1993). In addition, among persons diagnosed
with congestive heart failure (CHF), better marital functioning (i.e., higher
satisfaction, more marital routines, more useful discussions, and greater positivity
expressed during disagreements) was associated with higher 4-year survival rates,
controlling for CHF severity (Coyne, Rohrbaugh, Shoham, Sonnega, Nicklas, &
Cranford, 2001). Finally, in a study of mild hypertension sufferers, Baker and
colleagues found that marital adjustment was associated with better coronary health
(measured by left ventricular mass; the weight of the chamber primarily responsible
for the pumping of the heart). More importantly, spousal contact had an influence on
3-year diastolic blood pressure (DBP), but the effect was moderated by marital
adjustment. Spousal contact was associated with a decrease in DBP among well-
adjusted couples, whereas the reverse was true for poorly adjusted couples (Baker et.
al., 2000). In other words, interacting with one’s partner promotes cardiovascular
health only if one has a good relationship.
The main effect model, on the other hand, suggests that social support
promotes health notwithstanding stress levels. In other words, social support is
always beneficial, whether or not stress is present. Consistent with this model, close
others may provide individuals with the opportunity to engage in positive social
interaction, which can produce positive affect, and allow for the expression of
positive emotion. In an extensive review, Lyubomirsky and colleagues describe
numerous benefits of positive affect, including benefits to social life and health
(Lyubomirsky, King, & Diener, 2005). Frequent positive affect is associated with
longevity, fewer colds and injuries, less pain, less frequent disease relapse, and, for
some diseases, decreased severity (Pressman & Cohen, 2005). Furthermore, positive
emotional expression has been linked to enhanced well-being (Harker & Keltner,
2001) and longevity (Danner, Snowden, & Friesen, 2001), presumably because of the
positive social interactions one experiences across the lifespan. Positive emotions also
appear to have an “undoing effect” on negative emotions (Fredrickson, 2000).
Positive emotional experiences facilitated participants’ return to cardiovascular
baseline following a negative mood induction, thus undoing the cardiovascular
consequences of negative emotion (Fredrickson & Levenson, 1998; Tugade &
Fredrickson, 2004). Perhaps when positive emotions are socially shared, partners aid
in the undoing of negative affect, contributing to better cardiovascular function.
In addition to fostering positive affective experiences, close others may
cultivate health through the provision of instrumental benefits. Providing caregiving,
loaning money, or accompanying an individual to a doctor’s appointment are just a
few examples of the many welfare-promoting forms of tangible support that friends
may offer. Furthermore, relationship partners may instrumentally promote each
other’s welfare by raising each other’s awareness of health-relevant information,
encouraging healthy lifestyle choices, and motivating each other to care for
themselves (Kiecolt-Glaser & Newton, 2001.
Empirical evidence exists in support of both models, and methodological
issues surrounding the operationalization of social support have been identified as a
source of divergent results. In a review of studies testing one or both models, Cohen
and Wills (1985) concluded that when social support was measured structurally (e.g.,
number of social contacts), the main effect model typically received support.
However, when social support was measured functionally (e.g., appraisal of
interpersonal resources), the stress-buffering hypothesis was usually supported (see
also Cohen, 1991). It may be noteworthy to point out that few people are stress-free
(Almeida & Horn, 2004), thus making it difficult to distinguish main effects from
stress buffering effects in the absence of a stress-fee sample. The position adopted
here is consistent with Hartup and Stevens (1997) thesis; close relationship partners
are apt to provide benefits in good times and bad (i.e., both main and buffering effects
are likely to occur). The proposed study contains structural and functional measures
of social life, and accounting for both will provide a more comprehensive
understanding of the links between social relationships and health.
Taken together these findings suggest social factors are influential
contributors to psychological adjustment, and to the functioning of the major
physiological systems important to long-term health, morbidity, and mortality risk
(McEwen, 1998). Longitudinally examining the connection between social
engagement and health and well-being outcomes has the potential to reveal important
information about this relationship. A strained social life could act as a chronic
stressor, contributing to poor psychological adjustment, and long-term health risk
(McEwen, 1998). Although the body may be particularly resilient in early adulthood,
the negative repercussions associated with lacking, or having poor social relationships
throughout adulthood may begin to manifest themselves more clearly during midlife,
when health begins to deteriorate. Conversely, the benefits associated with an
enriched social life may begin to show their value at the same stage, when resilience
is on the decline. Although testing multiple birth cohorts would be necessary to make
any conclusive statements to that effect, if the current hypotheses are supported it
lends evidence to such a theory.
A Lifespan Approach
Aging and health. According to the Trends in Health and Aging initiative
organized by the Centers for Disease Control, the prevalence of diabetes,
hypertension, and coronary heart disease each triple among individuals age 45 and
over (National Center for Health Statistics, 2002; 2005). This increase in illness and
disease prevalence is likely due to decreased resiliency that accompanies aging.
Across the years, repeated exposure to physical and psychological stressors weakens
the body’s stress response efficiency, and it becomes increasingly difficult to return to
baseline levels of physiological functioning (Sapolsky, 1992; Sterling & Eyer, 1981).
Decreased resilience is a corollary of the glucocorticoid cascade hypothesis of
aging (Sapolsky, 1992). In a complex set of processes, stress activates the HPA axis
to ultimately produce glucocorticoid hormones (e.g., cortisol). Activation of the HPA
axis is generally adaptive because it mobilizes physical resources that facilitate
coping with the demands of stressors. However, glucocorticoids have the side effect
of causing hippocampal senescence. When the hippocampus deteriorates it is
prevented from serving the important neuroendocrine function of inhibiting the HPA
axis from further glucocorticoid production (thus making it a cycle or “cascade” of
aging surrounding the regulation of stress hormone secretion; McEwen, 1998;
Sapolsky, 1992). Chronic stress, and/or failure to inhibit cortisol production may
result in a flattening of the natural diurnal cortisol pattern (which typically peaks
about 45 minutes after awakening, and declines over the course of the day), or an
elevation in its mean level. Such dysregulation signals risk for disease because
cortisol suppresses the immune system, and may advance the development chronic
diseases such as diabetes and hypertension (McEwen, 1998; Sapolsky, 1990; 1996).
Negative health consequences become progressively more apparent as individuals age
and once-efficient homeostatic mechanisms are reset to less effective levels of
functioning (Sterling & Eyer, 1981). By mid and later adulthood the glucocorticoid
cascade may be well underway, and resilience may already be diminished (Sapolsky,
1992). Research on loneliness supports this idea; physiological costs to the
cardiovascular and neuroendocrine systems associated with loneliness heighten with
age (Hawkley & Cacioppo, 2007).
HPA axis activity is only one of the body’s responses to stress. In broader
terms, exposure to stress triggers the allostatic mechanism, which brings about
changes in all of the body’s physiological systems (e.g., neuroendocrine,
cardiovascular, and immune) in order to protect the body from harm (McEwen,
1998), and return the body to basal levels of functioning. Over time, and under certain
conditions (such as frequent stress, or difficulty adapting to repeated stressors), the
“cumulative biological burden exacted on the body through attempts to adapt to life’s
demands” may have a lasting negative effect, termed allostatic load (McEwen, 1998;
McEwen & Steller, 1993; Seeman, McEwen, Rowe, & Singer, 2001, p. 4770).
HPA axis activation is one of the most common allostatic responses
(McEwen, 1998), and dysregulation is considered an early marker of allostatic load
(Abercrombie, Giese-Davis, Sephton, Epel, Turner-Cobb, & Spiegel, 2004). More
substantively, the HPA axis is particularly responsive to socially situated stressors
(Dickerson & Kemeny, 2004), presumably because group belonging and social
acceptance have strong functional significance to survival (Baumeister & Leary,
1995). Laboratory stressor that contained social-evaluative threats (e.g. public
speaking) produced significantly larger cortisol effects than stressors that did not
contain social-evaluative threats (e.g., noise exposure; Dickerson & Kemeny, 2004).
Thus, cortisol level and rhythm will be focal outcomes of interest in the proposed
Early adulthood to midlife. The glucocorticoid cascade hypothesis seems to
suggest that stress in early and mid adulthood may be particularly important to later
life health. In a large-scale study of daily stress, the frequency of stressors was
highest in early adulthood (age 25-39), declined slightly into mid-adulthood (age 40-
59), and declined more sharply into older adulthood (age 60-74; Almeida, 2005;
Almeida & Horn, 2004; Birditt, Fingerman, & Almeida, 2005). Paradoxically, there is
relatively little research on the phenomena transpiring during adulthood that may
contribute to stress response efficiency. Skimming the contents of most human
development textbooks reveals that the lion’s share of attention surrounds changes
that occur through childhood and into adolescence. There is typically at least an entire
chapter devoted to old age; however, the section on middle adulthood usually
amounts to no more than a few pages. Believing that more than a few paragraphs
worth of interesting and important things are happening to people between the ages of
18 and 60 has energized an effort to study adulthood more closely. Large-scale
ventures such as the Midlife in the United States (MIDUS) study of more than 7000
American adults aged 25 through 74, and the Wisconsin Longitudinal Study (WLS)
of over 10,000 graduates from Wisconsin high schools, have begun to make headway
in filling this void. The current research is an attempt to contribute to that effort.
Adopting a lifespan approach is particularly useful for investigating the
relationship between social life and health. Not only is midlife a time when health
problems begin to emerge more rapidly, a growing and provocative literature
implicates experiences in early adulthood as a predictor of important outcomes in
later life. Findings from large epidemiological surveys, such as the MIDUS study,
show links between early life trauma (e.g., physical abuse, parental divorce) and
poorer mental and physical health in midlife (Goodwin, Hoven, Murison, & Hotopf,
2003; Maier & Lachman, 2000). In addition, positive emotional expression in early
adulthood predicts enhanced health and well-being outcomes decades later. In the so-
called “Nun Study,” a study of aging and Alzheimer’s among hundreds of nuns
belonging to the School Sisters of Notre Dame, a sub-sample of nuns were asked to
submit an autobiography prior to taking final vows in the early 1930s when the nuns
were approximately 22 years old. These handwritten autobiographies were coded for
emotional content (positive, negative, and neutral), and the nuns were surveyed
through the 1990s for mortality. Controlling for age and education, positive emotional
expression in the handwritten autobiographies was associated with decreased
mortality risk nearly 60 years later (Danner, Snowden & Friesen, 2001).
In a conceptually similar study, women’s college yearbook photos from the
late 1950s were coded for positive emotional expression. Women who displayed
relatively more “Duchenne” (genuine) smiles in their yearbook photos were
psychologically better adjusted, were more likely to become and stay married, and
were more likely to be satisfied in their marital relationship up to 30 years later. Of
importance, those results were not an artifact of physical attractiveness (Harker &
Keltner, 2001). Longitudinal effects such as these are believed to be mediated through
social channels. For example, emotional expressivity is hypothesized to shape social
experiences, which in turn contribute to health (Harker & Keltner, 2001; Keltner,
2003). In fact, in the yearbook study, women who displayed greater positive affect
were judged more favorably by independent observers, and were expected to be more
rewarding interaction partners (Harker & Keltner, 2001). However, little empirical
evidence exists to verify these mechanisms.
Social relationships as the mechanism. Social relationships provide a
theoretically plausible mechanism for linking early adulthood experiences to midlife
outcomes. At any stage in life social relationships have a powerful impact on human
development, influencing cognition, emotion, and the formation and maintenance of
subsequent relationships (Reis, Collins, & Berscheid, 2000). Object relations theorists
such as Sullivan (1953) and Bowlby (1969) argued that early social relationships give
rise to affective-cognitive structures related to self and social relations. Those
structures are believed to guide expectations about the manner in which interactions
between self and other should unfold throughout all of life (Bowlby, 1969; 1979;
Bretherton & Munholland, 1999).
Several lifespan development theorists concur that early adulthood marks the
point at which the formation of long-term adult friendships and love relationships
become the primary concern. Erikson (1959) was among the first to speculate about
development in adulthood, identifying the formation of close relationships as the
pivotal task of young adulthood. His psychosocial crisis of intimacy versus isolation
surrounds the acquisition, development, and maintenance of intimate social
relationships with friends and romantic partners during the late teens and early
twenties. Others have since argued for an analogous theme at this stage in life. For
example, Neugarten’s research has similarly identified early adulthood as the
prescriptive time on the “social clock” for emphasizing close intimate relationships,
such as marriage (Neugarten, Moore, & Lowe, 1965); White (1966) argued that a
developmental growth tendency in adulthood involved attending to relationships with
important others; and Levinson (1978) suggested that finding a spouse was one of the
key features of entering the adult world. Successfully establishing intimacy in early
adulthood fosters the ability to enter and maintain close relationships throughout the
remainder of adult life. For example, college students who achieved a higher level of
intimacy had greater knowledge and understanding of a same- or opposite-sex friend
(Orlofsky, 1976), and were more in touch with their own emotional experiences
(Orlofsky & Ginsberg, 1981), two qualities that would enhance their ability to form
and sustain close relationships.
Affirming the idea that there is something particularly important about the
relationships formed in late teens and early twenties, there is some evidence for
continuity in social relations throughout early adulthood (from college to age 30; Reis
et al., 1993), and even into old age (Field, as cited in Hartup & Stevens, 1997; Field,
1999; Roberto, 1997). Reis et al. (1993) found that although there were changes in the
nature of social interaction from the early twenties to age 30, intimacy levels
exhibited a substantial amount of stability across this time span. This suggests that the
quality of a person’s social ties is likely to persist over time, increasing the likelihood
of consistency in the benefits that will be afforded to individuals throughout life. If
individuals are surrounded with positive social ties when faced with various stressful
transitions occurring throughout adulthood (e.g., selecting a career, becoming self-
sufficient, buying a home, starting a family), those supportive others may repeatedly
facilitate adaptive psychological and physiological responses to stress.
Notwithstanding the continuity of relationship quality, the benefits gained
from good friends at any one point in time (e.g., psychological well being, coping
abilities, etc.) may be carried forth to future adaptations (Hartup & Stevens, 1997). In
other words, having an available and supportive network to aid one through a crisis
may provide an individual with certain tools that can be applied to future problematic
situations. Perhaps individuals embedded in extensive social networks of caring and
supportive friends will have repeated opportunities to develop adaptive coping
strategies that may be drawn upon over the life course (Hartup & Stevens, 1997).
Preliminary evidence indicates that a lifetime of positive social relation is associated
with reduced allostatic load among adults in their late 50s (Seeman, Singer, Ryff,
Love & Levy-Storms, 2002), suggesting that good relationships may indeed have a
cumulative positive effect on health. However, the sample is cross-sectional, and
reports are retrospective, two problems discussed in greater detail below.
In summary, models of relationships may shape the nature of intimacy
development in early adulthood, which sets the stage for the rest of adult social life.
If one has difficulty developing intimate relationships in early adulthood, or develops
an aversive pattern of social relations, this may act as a chronic stressor, wearing
down one’s resilience and increasing risk for disease (McEwen, 1998; Sapolsky,
1992). On the other hand, individuals who successfully navigate the intimacy task of
early adulthood have the potential to reap the psychosocial and physiological rewards
of good relationships for years to come.
The Current Research
Suggesting that social engagement may have a cumulative long-term effect on
health and well-being seems plausible based on the numerous studies reviewed above,
which concurrently link social factors with physiological and psychological
functioning. However, to better test the viability of such an argument, longitudinal
investigations are needed. An existing group of individuals who participated in detail-
rich social interaction research while they were undergraduates in college offers an
exceptional opportunity to longitudinally examine the long-term health and well
being correlates of early adult social life.
The necessity of longitudinal designs. Although longitudinal designs have
drawbacks of their own (they are expensive, they require extensive time and effort,
they are subject to attrition, and they possibly confound socioculturally significant
historal events with developmental changes) they offer at least three important
methodological advantages over cross-sectional designs. First, cross sectional studies
that show certain characteristics of social life to have similar means and correlates for
different age groups may not necessarily reflect true stability across time. A variety of
important changes may occur at the individual level, but change may be disguised if
individuals are changing in different directions and group means offer the only
comparison. For example, some individuals’ social life may worsen over time, while
others’ may improve, resulting in no overall mean difference. True change, and the
outcomes that may accompany it, is only revealed when same individuals are
followed across time. This is important for the current research because the
developmental journey does not end at age 18; a great number of transitions occur in
the lives of adults from age 18 to age 50. Transitions may include physical relocation,
finding a new social network, selecting a career, establishing financial independence,
buying a home, making a long-term romantic commitment (and possibly dissolving
that commitment), having children, watching grown children leave the nest, and
caring for elderly parents, just to name a few. Each of these relatively normative
events requires substantial adjustment, and is likely to bring about considerable
differences in social activity. Longitudinal research is able to illustrate individual
differences in the way people adjust to normative transitions, and relate those
differences to important later life outcomes.
Longitudinal investigations have the added benefit of reducing certain biases
associated with cross-sectional research designs. Cross-sectional studies must rely on
an individual’s retrospective aggregated impression of social life, the accuracy of
which cannot be determined. When reporting on events in the past, individuals tend to
rely on current circumstances to make heuristic-driven inferences about past behavior
and experience (Schwarz, Groves, & Schuman, 1998), and the retrospection bias
worsens with increased temporal distance from the recalled event (Skowronski, Betz,
Thompson, & Shannon, 1991). Furthermore, retrospective reports may be unduly
influenced by the participant’s mood at the time the measure is completed (Blaney,
1986). For example, in one study, individuals in a positive mood recalled fewer
negative life events, whereas individuals in a negative mood perceived social support
to be less available to them (Cohen, Towes, & Flocco, 1988). Thus, asking people to
retrospect about what social life was like in early adulthood produces a qualitatively
different type of evaluation. Although retrospective accounts are interesting in their
own right, longitudinal studies like this one, which obtain contemporaneous accounts
of social life, allow for a greater degree of accuracy by eliminating retrospection bias.
Finally, longitudinal investigations may help to shed light on causal
relationships. Undoubtedly, experimental manipulation is the standard criterion for
making cause and effect inferences. However, testing whether social life causes
disparities in long term health and well being is complicated because manipulating
individuals’ social lives on a large scale does not seem a viable (or reasonable)
option. When experimental manipulation is unsuitable, prospective longitudinal
studies may be the best strategy for exploring causal relationships (Cook & Campbell,
1979; Kenny, 1975; Taris, 2000). The design of this research is such that social life in
early adulthood was studied contemporaneously, and temporally precedes midlife
health and well-being outcomes. In addition, there is a theoretical rationale for the
hypothesized effects. If the data show the predicted effects, and if alternative
explanations are ruled out, the criteria for making a causal inference will be met
The existing sample. Three college-student samples from the University of
Rochester participated in event-contingent diary studies of social interaction between
1974 and 1980, and are currently between 49 and 52 years old (additional details
about the samples can be found in the Method section). The original studies were
aimed at uncovering sex differences in the patterns of social interaction among
college students (Wheeler & Nezlek, 1977), and the relationship between physical
attractiveness and social competence (Reis, Nezlek, & Wheeler, 1980; Reis, Wheeler,
Spiegel, Kernis, Nezlek, & Perri, 1982). These samples possess particularly desirable
characteristics for a longitudinal investigation of the correlates of early adult social
life because experience sampling methods were used to assess social life in early
adulthood. A weakness of the existing longitudinal research in this area is the one-
time assessments of social functioning using broad, global indicators. Such measures
have limited usefulness because they are vulnerable to the heuristics used in making
everyday prediction (Kahneman & Tversky, 1982), and they preclude identifying
which particular features of social life (e.g., frequency of interaction, number of
contacts, perceived interaction quality) may contribute to health. Experience
sampling methods have the ability to refine our assessment of social life.
Experience sampling methods. Experience sampling involves studying
ongoing social experience as it occurs in the natural ebb and flow of everyday life.
During the sampling period, participants complete a diary and report on the subject
matter of interest. Experiences are typically sampled using one of three protocols: at a
specified time interval (e.g., every evening before going to bed), in response to a
fixed or random signal (e.g., participants are alerted by a pager several times over the
course of a day), or following the occurrence of a specific event (e.g., each time a
social interaction occurs, as in the existing research).
The Rochester Interaction Record (RIR; Reis & Wheeler, 1991) is a
prominent experience-sampling instrument, and was used to assess early adulthood
social activity in the current research. The RIR (see Appendix A for a sample, but
note the earliest two samples completed a version that contained fewer questions,
namely intimacy and satisfaction as the only indicators of interaction quality) is an
event-contingent diary, which has participants complete a record describing pre-
selected characteristics of each social interaction lasting 10 minutes or more during
the designated diary period (typically 14 days). Investigators have successfully used
the RIR to examine an array of social interaction topics including conflict
(Pietromonaco & Feldman-Barrett, 1997), courtship (Milardo, Johnson, & Huston,
1983), friendship (Lydon, Jamieson, & Holmes, 1997), and intimacy (Reis et al.,
1993), among others.
Despite the fact that diaries are self-reports, an experience sampling approach
provides a more valid and reliable description of social life than that which is
obtained with single measurement surveys. Records are completed immediately
following each social interaction, lessening the delay between the occurrence and
report of events, and further reducing the amount of distortion associated with
aggregation and retrospection (Reis & Gable, 2000). Equally important, the RIR
contains diverse questions about the nature of each interaction, including both
objective (duration, number of people, sex composition) and subjective (intimacy,
satisfaction,) items. The variety of items allow for a more fine-grained analysis of
exactly which features of social life are predictive of later health and well being.
Although it would be desirable to use an experience sampling approach in the current
assessment, the resources and time commitment required for such an endeavor are
Results from the original studies indicate that reports of college social life
collected with the RIR are related to social and health variables. For example, lonely
men and women interacted with women less frequently, and perceived their
interactions with men and women to lack meaning, relative to non-lonely people
(Wheeler, Reis, & Nezlek, 1983). In addition, individuals with poor quality
interactions visited the health center more often (Reis et al., 1982).
A subset of 114 participants were recruited from the three college samples and
followed up in 1985 when participants were approximately 30 years old. Participants
once again completed event-contingent social interaction diaries for 14 days, as well
as global reports of social functioning, emotional well being, and self perceived
health. Results revealed that over time, opposite-sex socializing became more
frequent, whereas same- and mixed-sex socializing became less frequent. In addition,
social interactions became more intimate over time, but not more satisfying. The
onset of these changes occurred slightly earlier for women (during college) than men
(between college and age 30; Reis et al., 1993). Although there were changes in the
nature of interactions across time, Reis et al. (1993) also found marked stability in the
number of interactions, the amount of time spent interacting with others, and
particularly the quality of interactions from college to age 30. These findings lend
support to the idea that socialization in early adulthood influences the remainder of
adult social life. Despite the numerous transitions occurring between college and age
30, and the changes in social activity that they impose, individuals maintained
important characteristics of the social milieu.
The current assessment. The research reviewed thus far demonstrates that
social relationships play a powerful role in determining concurrent health and well
being, and suggests that they may also have similar long-term effects. Given the
notable stability in certain features of social life across early adulthood (Reis et al.,
1993), lacking, or have poor social relationships during that period may imply
problematic relationships across the remainder of adulthood, which may act as a
chronic stressor. However, having good social relationships at that time is likely to
facilitate healthy adaptation to the transitions occurring throughout adulthood. Thus,
social circumstances may have a cumulative impact on health and well being, which
would be evident in midlife indices of adjustment and health. The existing data
provide an ideal opportunity for investigating this possibility.
The proposed research employs a longitudinal design to investigate the
midlife social and emotional well being, and biological health correlates of social
activity during college. The use of RIR data to assess social activity across early
adulthood improves upon previous longitudinal research because it allows for the
examination of specific and distinct features of the social environment as predictors
of later life outcomes, and it is less vulnerable to retrospection biases. Moreover,
these detailed accounts of social activity were obtained during the period critical to
adult relationship development (Erikson, 1959; Neugarten et al., 1965), and not
Participants from the three previously described studies of college social
activity (between 49 and 52 years old at the time of the current assessment) were
recruited to participate in a follow-up study of health and well being. Midlife social
and emotional well-being was assessed with a series of internet-hosted
questionnaires. Health was measured with one objective (neuroendocrine function)
and several subjective (self report) indicators. A series of salivary cortisol samples
were collected over the course of an average day to capture HPA axis activity as one
indicator of physiological function, and self-reported health was assessed in the
Hypotheses. Support is sought for hypotheses in three domains of midlife
functioning: social well-being, emotional adjustment, and physical health (specific
measures of each are described in the Method section). Because structural (social
integration) and functional (relationship quality) measures of social life have both
been implicated in the outcomes of interest, quantity and quality relevant indicators of
social life will be tested as predictors for each hypothesis. Both types of variables are
provided by the RIR. Social integration will be operationalized as the average amount
of time spent socializing with others per day (time per day), and the average number
of different social interactions one has per day (number per day). Relationship quality
will be operationalized with ratings of intimacy and satisfaction for social
In addition to testing overall social activity, hypotheses will also be separately
evaluated for same-sex and opposite-sex social activity. The nature of social
engagement changes over time whereby individuals tend to engage less in same-sex
interactions and more in opposite-sex interactions (Reis et al, 1993), however, it is
unclear whether relative differences in the amount of same- versus opposite-sex
social interaction in college are associated with differential later outcomes. Testing
each separately will allow us to evaluate whether those differences exist. The
following hypotheses are offered.
Social well-being hypothesis. Structural and functional social engagement in
college will be positively associated with better social well-being in midlife (Figure
Emotional adjustment hypothesis. Structural and functional social
engagement in college will predict enhanced emotional adjustment in midlife (Figure
Health hypotheses. Structural and functional social engagement in college
will predict better health in midlife (Figure 1c). This hypothesis will be tested
separately for self-reported (subjective) health and neuroendocrine function.
Chapter 2: Method
One hundred thirty three (59.9%) participants age 48-52 (Mage = 49.28; 73
female, 59 male, 1 transgender) were recruited for the current research from a pool of
222 eligible adults who had previously participated in one of three event-contingent
diary studies while they were undergraduates at the University of Rochester. The
remaining 89 potential participants could not be located (n = 42, 18.9%), explicitly (n
= 15, 6.8%) or tacitly (i.e., did not respond to multiple contact attempts) refused to
participate (n = 23, 10.4%), or were deceased (n = 9, 4.1%). Seventeen of the 133
participants (12.8%) refused to submit saliva samples, but provided survey data.
College data could not be recovered for four of the 133 current participants,
effectively resulting in a sample of 129 participants who provided both college and
Of the 133 current participants, 34 (25.6%) were recruited from college
sample I (original n = 58). These participants had engaged in one to two waves of a
ten-day diary study of sex differences in social interaction during the 1974 – 1975
academic year, when they were freshman (Wheeler & Nezlek, 1977). Forty-three
(32.3%) participants were recruited from college sample II (original n = 77). These
participants were also college freshman at the time of initial participation. They
completed between two and four waves (administered twice each in the fall and
spring semesters) of a ten-day diary study of physical attractiveness and social
competence in the 1976 – 1977 academic year (Reis, Nezlek, & Wheeler, 1980).
Lastly, 56 (42.1%) participants were recruited from college sample III (original n =
113).1 This sample consisted of seniors who participated in a fourteen-day diary study
of physical attractiveness and social competence in the 1979 – 1980 academic year
(Reis, Wheeler, Spiegel, Kernis, Nezlek, & Perri, 1982; see Table 2 for a summary).
Participants were mailed a recruitment invitation at their most recently
available mailing address. Current contact information was obtained based on
information provided at the last assessment, as well as information made available
through online locator services including the University of Rochester alumni
database, online phone books, Yahoo people search, www.zabasearch.com, and
Recruitment invitations requesting phone numbers and e-mail addresses were
mailed to potential participants. Return receipt of the contact sheet prompted a
telephone call to give an overview of the project, and discuss details of participation.
Those who agreed to participate were mailed a participation kit including consent
forms, a cover letter with complete instructions for participation, a saliva sampling kit
(described below), and a postage-paid, self-addressed return envelope. In addition,
participants were emailed links to the multi-part web survey of health and well-being
(described below), along with detailed instructions for completing the web surveys,
and an ID code to use in place of other identifying information. Participants were
instructed to return all materials to our laboratory on the day following saliva
collection. Upon receipt of the completed materials, participants received a fifty-
As discussed above, the RIR (Reis & Wheeler, 1991) includes questions that
assess interaction quantity, or structural social integration, as well as questions that
assess interaction quality, or functional social support. Two questions were selected
to represent interaction quantity (average time per day spent interacting and average
number per day of interactions), and two questions were selected to represent
interaction quality (average interaction intimacy, and average interaction satisfaction).
Time per day and number per day were significantly correlated for overall social
activity (r = .42, p < .001), same-sex social activity (r = .81, p < .001) and opposite-
sex social activity (r = .88, p < .001), thus time per day and number per day were
standardized and combined into three respective interaction quantity variables
(overall, same-sex, opposite-sex). Similarly, intimacy and satisfaction were
significantly correlated for overall social activity (r = .43, p < .001), same-sex social
activity (r = .41, p < .001), and opposite-sex social activity (r = .30, p < .001), and
were combined into three respective interaction quality variables (overall, same-sex,
Surveys. An Internet survey-hosting website (www.surveymonkey.com)
hosted a three-part assessment of social and emotional well-being, and self-reported
health, described in detail below. Participants logged into the survey with a unique
identifying code and completed the self-paced questionnaire at their convenience. The
majority of survey sections were completed in one sitting (93.1%) with an average
completion rate of 38 minutes per section.
Emotional adjustment measures. A set of emotional well-being measures was
carefully selected to reflect both the positive and negative dimensions of well-being,
which have recently been described as independent (Ryff et al., 2006). The measures
were intended to cover key emotional adjustment constructs with minimal
Psychological well being. A 54-item measure of psychological well-being
assessed six domains of wellness (autonomy, environmental mastery, self-acceptance,
positive relations with others, personal growth, and purpose in life) with 9 items per
subscale (Ryff & Keyes, 1995). Responses were rated on a 6-point scale ranging from
disagree strongly to agree strongly. Respective Cronbach’s alphas for the six
subscales are: .86, .81, .90, .91, .80, and .79.
Depression. The Center for Epidemiological Studies 20-item measure
assessed depression during the preceding week (CES-D; Radloff, 1977; ! = .90).
Responses were rated on a four-point scale ranging from rarely or none of the time
(less than 1 day) through most or all of the time (5-7 days). In addition, the 13-item
depression subscale of Symptom Checklist 90 (SCL-90; Derogatis, 1977) measured
depressive symptomatology during the preceding four weeks (! = .88). Responses
were recorded on a 5-point scale from not at all through extremely.
Anxiety. A 10-item subscale of the Symptom Checklist 90 (SCL-90;
Derogatis, 1977) measured anxiety during the preceding four weeks (! = .74).
Responses were recorded on a 5-point scale from not at all through extremely.
Stress. The 10-item Perceived Stress Scale (PSS; Cohen, Kamarck, &
Mermelstein, 1983) measured feelings of stress over the last month (! = .87).
Responses were made on a 5-point scale from never through very often.
Life satisfaction. Life satisfaction was measured with the 5-item Satisfaction
with Life Scale (SWLS; Diener, Emmons, Larsen, & Griffin, 1985). Responses were
made on a 7-point scale ranging from strongly disagree through strongly agree (! =
Loneliness. A 10-item version of the UCLA loneliness scale (UCLA-LS;
Russell, 1996) measured feelings of social isolation with responses given on a 4-point
scale from never to often (! = .89). The 10 items were the same ones that had been
used with one of Russell’s (1996) samples (teachers) during scale validation.
Mood. Positive and negative affect experienced during the preceding week
was assessed with the 18-item version of the Positive and Negative Affect Schedule
(PANAS-X; Watson & Clark, 1994). Eight positive and 10 negative emotions were
rated on a 5-point scale ranging from very slightly or not at all through extremely
(respective Cronbach’s alphas were .90 and .87).
Social well-being measures. A set of questionnaires was selected to assess
social adjustment at broad and specific levels. General questionnaires captured
overall level social integration, whereas detailed questionnaires assessed specific
types of relationships relevant to most individuals in midlife (including romantic
relationships, and closest friends).
Social integration. The Social Network Index (Cohen, Doyle, Skoner, Rabin,
& Gwaltney, 1997) assessed levels of social integration. Participants reported on the
number of different kinds of relationships they had, and the frequency with which
they had contact with each. Indices of social network size (i.e., number of different
social contacts), and social network diversity (i.e., number of social roles filled) were
derived from this measure.
Perceived social support. The Kessler Perceived Support Scale (KPSS;
Kessler, Kendler, Heath, Neale, & Eaves, 1992) measured perceptions of the
availability of tangible, emotional and esteem support from six sources: family (! =
.85), friends (! = .87), spouse/partner (! = .97), people from church/synagogue (! =
.97), people from neighborhood (! = .93), and professional colleagues (! = .94).
Ratings are made on a 4-point scale ranging from 1 not at all through a great deal.
Relationship quality with core group of friends. Participants were asked to
think about their core network of closest friends. This was described as “the people
they see and interact with most often, participate in activities with (e.g., dinner,
shopping, golf, etc.), and would include in the planning of a get-together.”
Participants completed four 6-item subscales (social intimacy, ! = .60; emotional
intimacy, ! = .79; intellectual intimacy, ! = .57; recreational intimacy, ! = .77) of the
Personal Assessment of Intimacy in Relationships (PAIR; Schaefer & Olson, 1981) to
measure how intimately connected they felt to this core group of relationship
partners. Ratings were made on a 9-point scale ranging from not at all true through
completely true. In addition, participants reported their closeness to this group using
the 1-item Inclusion of Other in the Self (IOS) scale, in which they select one of
seven pairs of increasingly overlapping circles that best represents their feelings of
closeness to the group (Aron, Aron, & Smollan, 1992).
Romantic relationship quality. Participants who reported being involved in a
primary romantic relationship were asked to describe the quality of that relationship.
Relationship quality measures included the IOS (Aron et al., 1992) to measure
closeness; the same four 6-item subscales of the PAIR (social intimacy, ! = .79;
emotional intimacy, ! = .92; intellectual intimacy, ! = .85; and recreational intimacy,
! = .81; Schaefer & Olson, 1981) used to measure intimacy with friends; the Couples
Satisfaction Inventory (CSI; Funk & Rogge, 2007) to measure relationship
satisfaction (! = .98); and the Perceived Partner Responsiveness scale (PPR;
Caprariello & Reis, 2006) to measure the extent to which participants believed that
their romantic partners understand, validate, and care for core aspects of the self (! =
Self perceived health measures. A selection of self-report measures of health
was chosen to assess participants’ feelings about their health status, health behaviors,
and any existing health conditions they may have. These measures were selected to
succinctly obtain as complete a picture of physical health status as possible. In
addition to the measures described below, participants reported height, weight,
cholesterol level, blood pressure levels, and how recently they had a medical
checkup, cholesterol test, and blood pressure check.
SF-36. Twenty-seven items from the Medical Outcomes Study Short Form 36
(SF-36; Ware & Sherbourne, 1992) assessed participants’ health in six domains:
physical functioning, physical limitations, emotional limitations, social functioning,
pain, and general health. Nine items related to depression were excluded to prevent
redundancy. Some items referred to experiences over the preceding four weeks, and
others assessed global perceptions without indicating a time frame.
Symptoms. The Cohen-Hoberman Inventory of Physical Symptoms (CHIPS;
Cohen & Hoberman, 1983) measured the extent to which participants were bothered
by 33 different symptoms of physical illness experienced over the past two weeks.
Ratings were made on a 5-point scale from not at all through extremely.
Prior conditions. Participants reported on whether they had ever been
diagnosed with any of 25 medical conditions (e.g., high cholesterol, high blood
pressure, diabetes, stroke, hayfever, etc.). Items were obtained from the National
Health Interview Survey (NHIS) conduced by the National Center for Health
Statistics of the Centers for Disease Control.
Sleep quality. The Pittsburgh Sleep Quality Index (PSQI; Buysse, Reynolds,
Monk, Berman, & Kupfer, 1989) assessed sleep quality. Participants responded to
questions about their sleep over the last month, spanning 7 categories including sleep
quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances,
use of sleeping medication, and daytime dysfunction.
Other measures. Participants completed the following additional self-report
measures in the online survey.
Demographic information. Participants reported on several demographic
variables including age, gender, race, ethnicity, education, and subjective
socioeconomic status (Adler, Epel, Castellazzo, & Ickovics, 2000), and financial
Personality. A 50-item measure of the Big-Five personality traits
(surgency/extraversion, ! = .88; agreeableness, ! = .85; conscientiousness, ! = .86;
emotional stability/neuroticism, ! = .81; intellect/imagination, ! = .84; Goldberg et
al., 2006) assessed participants’ personality with 10 items per subscale. Responses
were rated on a 9-point scale ranging from extremely inaccurate to extremely
Neuroendocrine function. To determine diurnal cortisol rhythm, participants
provided five saliva samples during a single sampling day using a protocol developed
by Kirschbaum and colleagues (Kirschbaum, Kudielka, Gaab, Schommer, &
Hellhammer, 1999; Schmidt-Reinwald, Preessner, Hellhammer, Federenko, Rohleder,
Schumeyer, & Kirschbaum, 1999; Wust, Federenko, Hellhammer, & Kirschbaum,
2000) and the John D. and Catherine T. MacArthur Research Network on
Socioeconomic Status and Health (Stewart & Seeman, 2000).
The number and timing of saliva samples to be collected was based on
guidelines determined at a meeting of the MacArthur Research Network on
Socioeconomic Status and Health held at Rockefeller University in December 1999
(Stewart & Seeman, 2000). The meeting participants were internationally recognized
researchers investigating diurnal cortisol measures. They agreed on a strategy of a
one-day, 6-sample protocol for establishing changes in magnitude and rhythm of
cortisol in study populations. We collected 5 of the 6 samples, dispensing with the
bedtime sample. Collection times were: (1) awakening (before getting out of bed); (2)
45 min after awakening; (3) 2.5 h after awakening; (4) 8 h after awakening; and (5)
12 h after awakening.
Saliva was collected using the Salivette collection system (Sarstedt,
Germany). Participants were instructed not to brush their teeth for at least 30 min
prior to any of the five saliva samplings to avoid possible contamination of samples
with blood from micro injuries, and were asked to refrain from brushing their teeth
until after the second sample of the day (see below) had been collected. Participants
were also asked to refrain from eating for 10 min prior to saliva sampling; particularly
foods low in pH (i.e., fruit juice). At the designated times, participants gently chewed
on the cotton Salivette swab for approximately 1 min. This generally resulted in a
saliva sample of 0.5-1 ml. Following collection, participants were asked to refrigerate
the samples, until mailing the entire set to our laboratory the next day.
A number of considerations are important for analyzing cortisol data. The
MacArthur Network established 6 categories of factors that have important
implications for variation in cortisol rhythm including (1) individual characteristics
(e.g., age and gender); (2) menstrual cycle, contraceptive use, medication use; (3)
general health, especially infectious disease, which may be associated with flattening
of the diurnal rhythm; (4) food intake, exercise, wake-up time; (5)
psychological/psychosocial factors; and (6) alcohol/smoking. Participants maintained
a log on the sampling day, and provided information on five of the six categories,
including food intake, smoking/alcohol usage, medications taken, exercise,
awakening time, and hours of sleep. The daily log did not contain measures of
psychological/psychosocial factors, which were assessed in the larger online
Sample storage and shipping. Samples were mailed to the laboratory at room
temperature. When samples arrived, they were frozen at -20°C to precipitate mucins.
Samples were assayed in bulk, with all samples from a given subject run in the same
batch. At the time of assay, samples were thawed completely, vortexed, and
centrifuged at 1500g for 5 min. The clear saliva was pipetted into appropriate wells of
the assay plate.
Cortisol Assay. All samples were assayed for salivary cortisol using a HS-
cortisol high sensitivity enzyme immunoassay kit (Salimetrics, PA). The test has a
lower limit of sensitivity of .007 !l/dl, range of sensitivity from .007 to 1.2 !g/dl, and
average intra-and inter-assay coefficients of variation 4.13% and 8.89%, respectively.
The correlation between saliva and serum is highly significant, r (17)=0.96, p<.0001.
The kit contains a built-in pH indicator that warns the user of acidic or basic samples.
As noted below, samples with an unusually high pH were excluded from analyses.
Chapter 3: Results
Participant attrition and comparability. Independent samples t-tests
compared the 129 participants in the current wave of data collection to 80 of the 89
eligible individuals from the original samples who did not participate in the current
study (9 non-participants were missing college data). There were no significant
differences in the college interaction quantity or quality variables for the two groups
(all ts < -1.50, ns; see Table 3), suggesting that recruited participants did not differ
from the unrecruited participants in any meaningful way in college social activity.
Two additional ANOVAs tested whether 1) sex, and 2) original sample from
which participants were recruited (I, II, or III, as described above), interacted with
current participation status. A 2 (Sex: male, female)2 x 2 (Participation Status:
recruited, unrecruited) ANOVA on 18 measures of college social activity revealed no
main effects of participation status, and no Sex X Participation Status interaction
reached significance. Similarly, a 3 (Original Sample: I, II, II) x 2 (Participation
Status: recruited, unrecruited) on the same 18 measures of college social activity
revealed that only 1 (5.6%) Original Sample X Participation Status interaction
reached significance. Figure 2 shows that currently recruited participants from sample
I had fewer opposite-sex interactions in college (M = 1.33) than unrecruited
participants from sample I (M = 1.98), whereas the reverse was true for sample II
(recruited M = 1.76, unrecruited M = .107), and there were no differences in the
number of opposite-sex interactions between recruited (M = 2.00) and unrecurited (M
= 1.79) participants from sample III, F(2, 201) = 3.68, p < .05.
Although there were sex differences in the social activity variables, the null
findings involving current participation status suggest that recruited participants did
not differ from unrecruited participants in any meaningful way on college social
An additional set of independent samples t-tests compared the college social
activity of unrecriuted participants who expressly declined (n = 15, but college data
were missing for two) versus those who tacitly refused by non-response to repeated
contact attempts (n = 23, but college data were missing for two). These two groups
did not differ in quantity or quality of overall, same-, or opposite-sex college social
activity (all ts < 1.36, all ps > .18). Moreover, neither of these sub-groups of
unrecruited participants differed from recruited participants on any of the college
social activity predictors (all ts < 0.89, all ps > .38). Thus, differences in the
willingness to respond to this outreach attempt did not vary as a function of early
adult social life.
Outcomes. Multiple measures assessed three broad categories of midlife
outcomes: social well-being, emotional adjustment, and self-reported physical health.
For parsimony and ease of interpretability, principal components analysis was
undertaken to reduce the number of outcomes to a smaller number of summary
variables representing each broader category. Principal components were accepted
based on three criteria: 1) Kaiser's criterion - that each principal component’s
eigenvalue is greater than 1; 2) scree test - a clear leveling-off of eigenvalues after the
last interpretable factor with an eigenvalue greater than one; and 3) interpretability.
To account for the differential scaling of individual measures, z-scores were
computed before combining measures to create unit-weighted composite factor
Social well-being measures. A principal components analysis with oblimin
rotation3 reduced the 10 measures of social relationship quality (social network
diversity; social network size; perceived support, closeness and intimacy with core
friend group; and perceived support, closeness, intimacy, satisfaction, and perceived
partner responsiveness from romantic partner for those participants currently
involved in a romantic relationship) to three principal components (see Table 4 for
factor loadings, eigenvalues, and communalities). The first factor, containing four
measures assessing romantic relationship quality (perceived support, closeness,
satisfaction, and perceived partner responsiveness for romantic partners), accounted
for 44.0% of variance. Factor two consisted of three scales assessing relationship
quality with core friend group (perceived support, closeness, and intimacy), and
accounted for 22.9% of variance. The third component, capturing social network,
contained two items (social network diversity, and social network size), accounting
for 14.5% of variance. Intimacy with romantic partner did not load strongly enough
with the other romantic relationship quality variables, instead loading on the
friendship quality factor. However, because of its lack of theoretical fit with the
friendship variables, this variable was not included on either factor, but was retained
for separate analysis. Three social well-being summary variables were created:
romantic relationship quality, friendship quality, and social network.
Emotional adjustment measures. A principal components analysis with
oblimin rotation reduced 14 measures of emotional well-being (satisfaction with life,
loneliness, perceived stress, anxiety, two measures of depression, positive and
negative affect, and the six subscales of Ryff’s psychological well-being measure:
personal growth, purpose in life, self acceptance, autonomy, environmental mastery,
and positive relations with others) to three principal components. Table 5 reports
factor loadings, eigenvalues, and communalities. The first factor, accounting for
52.0% of variance, contained items associated with positive emotional adjustment
(life satisfaction, self-acceptance, positive affect, positive relations with others, and
the absence of loneliness). The second factor, accounting for 11.2% of variance,
contained items associated with impoverished emotional adjustment (negative affect,
both measures of depression, anxiety, and perceived stress). Factor three, accounting
for 7.3% of variance, contained items related to self-actualization, or maximizing
one’s inner potential (personal growth, purpose in life, and autonomy). The
environmental mastery subscale loaded equally on all three factors, and was omitted
from any factor because of its inconsistency. Three emotional adjustment summary
variables were created: positive affectivity, negative affectivity, and self-actualization.
Physical health measures. A third principal components analysis reduced
nine self-reported indicators of physical health (cholesterol level, systolic blood
pressure, diastolic blood pressure, body mass index, history of cardiac diagnoses
[proportion of 7 conditions: high cholesterol, high blood pressure, coronary heart
disease, angina, heart attack, other heart condition, and stroke], history of respiratory
diagnoses [proportion of 4 conditions: emphysema, asthma, sinusitis, and chronic
bronchitis], history of endocrine diagnoses [proportion of 7 conditions: diabetes, pre-
diabetes, impaired fasting glucose, borderline diabetes, high blood sugar, and family
history of diabetes], symptomatology, and a combination of the 4 physical health
related subscales of the SF-36 [physical functioning, the extent to which physical
health limits activity, pain, and a global measure of self-perceived healthfulness]) to
three principal components. Factor loadings, eigenvalues, and communalities are
reported in Table 6. Three variables loaded on the first component (symptomatology,
SF-36 physical health, and respiratory diagnoses), accounting for 25.7% of variance.
The second factor also contained 3 variables (cardiac diagnoses, systolic blood
pressure, and diastolic blood pressure), accounting for 16.1% of variance. The third
and final factor contained 3 variables (cholesterol, body mass index, and endocrine
diagnoses), accounting for 12.5% of variance. Cardiac diagnoses loaded on both the
second and third factors, but had a stronger factor loading on component two, and
was placed there with the two blood pressure items. Three physical health summary
variables were created: daily symptomatology, cardiac health I, and cardiac health II.
Neuroendocrine function. One hundred sixteen of 133 current participants
provided between 3 and 5 saliva samples over the course of one sampling day for a
total of 571 saliva samples. The majority of participants provided all five samples (n
= 109, 94.0%), five participants missed one sample, and two participants missed two
samples. The samples were assayed for cortisol, as described above. Twenty-one
cortisol values were omitted because the pH of the sample was outside of an
acceptable range, and the cortisol assay thus could not be trusted. An additional 21
values were omitted because they were more than two standard deviations away from
the mean cortisol level for that sampling time, and were also considered unreliable.
The multiple cortisol measurements taken over the course of one day were
combined into two cortisol variables (Pruessner, Kirschbaum, Meinlschmid, and
Hellhammer, 2003). One variable captured total cortisol concentration over the course
of the day (area under the curve with respect to ground; AUCg). The second variable
removes the area between zero and the first cortisol value of the day to capture
changes in cortisol over the course of the day with respect to a person’s starting
cortisol value (area under the curve with respect to increase4; AUCi). Cortisol
concentration (AUCg) and change (AUCi) were negatively correlated (r = -.24, p <
.05); a greater change (i.e., steeper decline in cortisol values) over the course of the
day was associated with a lower overall cortisol concentration.
For all analyses involving these two cortisol variables, an additional set of
analyses were conducted that also controlled for the previously described factors
reputed to influence cortisol levels: sex, age, hours of sleep the previous night,
whether the participant exercised, smoked, or experienced any symptoms of illness on
the sampling day, and oral contraceptive use (for females only). Although only two of
these factors were significantly related to cortisol (participants who got more sleep on
the night prior to sampling had lower cortisol concentration, AUCg r = -.28, p < .01;
and participants who smoked on the sampling day had a smaller change in cortisol
values over the course of the day (AUCi r = -.21, p < .05), they were all included as
controls, and negligibly altered the results obtained without their inclusion. The
effects presented thus do not include the aforementioned controls.
Sex, Original Sample, and Freshman Versus Senior Status
A set of MANOVAs examined differences in midlife outcomes as a function
of sex, original college sample, and their interaction. A contrast on the original
sample variable compared individuals who participated as freshman (samples I and
II) to those who participated as seniors (sample III). As previously noted, 26
individuals participated in both samples II and III, and were randomly assigned to one
of the two samples (see Footnote 1 for an explanation). There were no significant
main effects of original sample on any of the midlife social well-being (see Table 7),
emotional adjustment (see Table 8), or physical health (see Table 9) composites, or
cortisol concentration (see Table 10). There was a significant effect of the college
year contrast on the friendship quality composite; participants from the freshman
samples had better quality friendships at midlife (M = .12) than participants from the
senior sample (M = -.09; F[1,122] = 5.64, p < .05). The college year contrast did not
have a significant effect on any other midlife composite variable (see Tables 7-10).
The Original Sample X Sex interaction reached significance for two of the
physical health composite variables: daily symptomatology (F[2,126] = 4.65, p <
.05), and cardiac health II (F[2,68] = 5.85, p < .01). As shown in Figure 3, simple
effects tests revealed that men and women from samples I and III did not differ
significantly from each other in daily symptomatology (sample I means: men = -.10,
women = -.16, t < 1, ns; sample III means: men = .00, women = -.06, t < 1 ,
ns). However, women from sample II reported significantly more symptoms (M =
.44) than men from sample II (M = -.36; t = -3.48, p < .01). For cardiac health II,
as shown in Figure 4, simple effects revealed that women from sample I had
significantly more cardiac health problems (M = .88) than men from sample I (M = -
.32; t = -4.01, p < .01), however the sex difference in cardiac health II was not
significant for men and women from samples II (men M = .25, women M = -.08; t
= 1.04, ns), or III (men M = -.04, women M = -.01; t < 1, ns).
The College Year X Sex interaction reached significance for one of the
emotional adjustment composite variables, positive affectivity, F(1,120) = 5.05, p <
.05. Simple effects tests show there was no difference in positive affectivity for men
(M = -.16) and women (M = .07) who participated as freshman (t < 1, ns),
however women who participated as seniors reported significantly more midlife
positive affectivity (M = .35) than men who participated as seniors (M = -.32; t =
-3.19, p < 01) (see Figure 5).
Eleven of 192 (5.7%) tested main and/or interactive effects involving original
college sample and college year emerged as significant. The largely null findings
involving these two variables suggest that original college sample and freshman
versus senior status did not have a reliable or consistent impact on midlife social,
emotional, health, or neuroendocrine outcomes, and these two variables are not
further discussed. Sex main effects emerged on several midlife variables (also
included in Tables 7-10), and are explored below when they moderate the key
analyses of interest.
Pearson correlations were conducted to examine relationships among midlife
outcomes. Table 11 contains a matrix of correlations among social well-being,
emotional adjustment, and physical health composite scores, which are described
below. Correlations between composite scores and the individual measures that were
used to create each composite are contained in Table 12 (social well-being and
emotional adjustment individual measures) and Table 13 (physical health and
neuroendocrine individual measures).
Correlations revealed that, at midlife, those who had better social lives were
also better emotionally adjusted. Adults with better friendships and romantic
relationships had less negative affectivity in midlife (respective rs = -.40, -.36, both
ps < .001), and more positive affectivity (respective rs = .62, .47, both ps < .001) and
self-actualization (respective rs = .28, .27, both ps < .001). A similar pattern emerged
between midlife social network and emotional adjustment, though the effect was
somewhat weaker for negative affectivity outcomes (respective correlations between
social network and negative affectivity, positive affectivity, and self actualization: r =
-.16, p < .08, r = .38, p < .01, r = .22, p < .01)
Social relationship quality was generally unrelated to physical health.
Romantic relationship quality and social network were not significantly correlated
with daily symptomatology or either cardiac health composite (all "r"s < .14, all ps >
.17). Friendship quality was not significantly related to daily symptomatology (r = -
.05, ns), or the second cardiac health composite (r = -.12, ns), but was negatively
correlated with the first cardiac health composite (consisting of prior cardiac
diagnoses, and systolic and diastolic blood pressure; r = -.27, p < .01). In other words,
individuals who have better social relationships at midlife also have better cardiac
health at midlife.
Midlife emotional adjustment was not consistently related to midlife health
outcomes. Negative affectivity was associated with greater daily symptomatology (r
= .29, p < .01), but unrelated to either cardiac health composite (both"r"s < .10, both
ps > .32). Positive affectivity was unrelated to daily symptomatology or the second
cardiac health composite (both "r"s < .12, both ps > .18), but was negatively
correlated with the first cardiac health composite (r = -.20, p < .05). Participants
experiencing more positive emotional adjustment reported better cardiac health. Self-
actualization was unrelated to any of the health composites (all "r"s < .12, all ps >
Midlife social relationship quality was not significantly related to cortisol
concentration (AUCg) or change (AUCi) for romantic relationship quality (respective
rs = .08, and -.07, ns), friendship quality (respective rs = -.06, and -.03 ns), or social
network (respective rs = .13, and -.03, ns). Emotional adjustment was also unrelated
to neuroendocrine function; correlations with negative affectivity, positive affectivity,
and self-actualization did not reach significance for cortisol concentration (respective
rs with AUCg= -.02, .03, .13, all ps > .19) or cortisol change (respective rs with
AUCi= .04, -.05, .02, all p’s > .19). Daily symptomatology and the first cardiac health
composite were not significantly correlated with cortisol concentration (respective rs
= -.05, .03, both ps > .60), but the second set of cardiac health outcomes was
marginally associated with higher cortisol concentration (r = .19, p = .06).
Participants who reported poorer cardiac health (higher cholesterol, higher BMI, and
a greater proportion of endocrine diagnoses) had higher mean levels of cortisol. None
of the physical health composites were significantly correlated with cortisol change
(daily symptomatology r = .06, cardiac health I r = .13, cardiac health II r = -.11, all
ps > .25).
Sex differences in concurrent correlations. Table 14 contains correlations
among the composite variables separately for men (below the diagonal) and women
(above the diagonal), revealing the general pattern of concurrent relationships
reported above to be highly similar for the two sexes. A few notable sex differences
emerged, and when they did, for the most part concurrent correlations between social
well-being and physical and emotional outcomes were stronger for men than for
women. The correlation between social network and self-actualization (r = .22, p <
.05) was stronger for men (r = .39, p < .01), but was not significant for women (r =
.10). The correlation between friendship quality and cardiac health I (r = -.27, p <
.01) was also predominantly driven by men (r = -.25, p < .10); the relationship
between these two variables was not significant for women (r = -.06). Although the
overall relationship between friendship quality and daily symptomatology was not
significant, and the same was true for men (r = -.01) women with better friendships
had marginally lower daily symptomatology (r = -.22, p < .10).
Longitudinal Prediction of Midlife Outcomes
Analysis strategy. Each midlife summary outcome variable was
simultaneously regressed onto the two primary college social activity composite
predictors: interaction quantity (the proxy for structural integration) and interaction
quality (the proxy for functional support). Sex, and the interaction between sex and
each predictor were also included in a second, incremental model.5 Analyses were
conducted for overall college social activity, as well as separately for same-sex and
opposite-sex college social activity. The text reports effects from the first model. As
mentioned above, sex main effects emerged on several midlife outcomes. When sex
moderated the effect of interaction quantity or quality on an outcome variable, the
results are described below.6
Tables 15 through 24 (as indicated below) present regression coefficients
(standardized #s) for both models on the composite outcome variable in the first
column, as well as the effects of college interaction quantity and quality on each of
the individual outcome measures comprising the summary variable in the remaining
Romantic relationship quality. Analyses of midlife romantic relationship
quality variables were limited to those participants who indicated they were
“currently involved in an exclusive, committed, romantic relationship” (n = 113).
Prior to examining romantic relationship quality, romantic relationship status was
tested as an outcome, and was not significantly predicted by overall college social
activity (quantity # = .02, ns; quality # = -.05, ns), same-sex college social activity
(quantity # = .10, ns; quality # = -.01, ns), or opposite-sex college social activity
(quantity # = -.07, ns; quality # = -.05, ns).
College social activity was not significantly related to midlife romantic
relationship quality (quantity # = .12, ns; quality # = -.04, ns). The same was true for
same-sex social activity (quantity # = .10, ns; quality # = -.04, ns) and opposite-sex
social activity (quantity # = .04, ns; quality # = -.07, ns). See Table 15 for a complete
summary of romantic relationship quality effects.
Although sex did not moderate the effects of college interaction quantity or
quality on the midlife romantic relationship quality composite for overall, same- or
opposite-sex interactions, there was a significant Same-sex Quantity X Sex
interaction on variable in the romantic relationship quality composite: perceived
partner responsiveness (# = .23, p < .05). Calculation of simple slopes revealed that
the quantity of same-sex interaction one had in college was marginally positively
associated with midlife perceived partner responsiveness for women (# = .25, p =
.09), whereas the association was not significant, for men (# = -.19, ns; see Figure 6).
Because intimacy with romantic partner (PAIR-R) did not load with the other
romantic relationship quality measures in the principal components analysis, it was
examined on its own. Both frequency and quality of social activity in college were
positively associated with romantic relationship intimacy in midlife (quantity # = .24,
p < .05; quality # = .22, p < .05). In other words, participants with more frequent
social activity in college, and participants with higher quality social activity in
college, reported more intimate romantic relationships in midlife.
Friendship quality. Participants who had more frequent social activity in
college had better quality friendships in midlife (quantity # = .21, p < .05). Although
there was not a significant effect of social activity quantity on perceived support from
friends (# = .12, ns), participants with more frequent college social activity felt
marginally closer to their core friend group (# = .16, p < .10), and had significantly
greater intimacy with their friend group (# = .24, p < .01). The effect of college
interaction quantity on midlife friendship quality was not significant for same- or
opposite-sex college interactions (both #s < .13, ns).
College social interaction quality was also associated with enhanced
friendship quality in midlife (# = .28, p < .01). In other words, participants who had
more intimate and satisfying interactions in college also reported better quality
friendships at age 50. This effect was consistent across all three measures making up
the friendship quality composite: perceived support # = .21, p < .05; closeness # =
.24, p < .01; intimacy # = .23, p < .01. College interaction quality was a significant
predictor of midlife friendship quality for both same-sex social activity (# = .31, p <
.01) as well as opposite-sex social activity (# = .19, p < .05), although across the
measures making up the friendship quality composite, the effects were consistently
stronger for same-sex college social activity quality (perceived support # = .24, p <
.01; closeness # = .26, p < .01; intimacy # = .27, p < .01) than for opposite-sex
college social activity quality (perceived support # = .08, ns; closeness # = .21, p <
.05; intimacy # = .17, p < .10). See Table 16 for complete results.
Although sex did not moderate the effect of interaction quantity or quality on
friendship quality for overall, same-, or opposite-sex social activity, there was a
significant Overall Interaction Quantity X Sex interaction on one of the measures in
the friendship quality composite: perceived support (# = .19, p < .05). Calculation of
simple slopes revealed that more frequent social activity in college was associated
with an increase in midlife perceived support from friends for women (# = .23, p <
.05), thought the relationship was not significant for men (# = -.16, ns; see Figure 7).
Social network. Quantity of social activity in college was positively associated
with social integration in midlife (# = .24, p < .01). People who were well connected
in college had larger (network size # = .24, p < .01) and more diverse (network
diversity # = .20, p < .05) social networks in midlife. The effect of college interaction
quantity on the social network composite was weaker for same-sex quantity (# = .15,
p < .10), and non-significant for opposite-sex quantity (# = .03, ns).
College interaction quality was unrelated to social integration in midlife, for
overall (# = .04, ns), same-sex (# = .04; ns), and opposite-sex social activity (# = .05,
ns). See Table 17 for a complete summary of effects.
Although sex did not significantly moderate the effects of social activity on
the social network composite variable for overall, same- or opposite-sex social
activity, there was a significant Same-sex Interaction Quality X Sex interaction on
one of the variables in the social network composite, network size (# = -.22, p < .05).
Calculation of simple slopes revealed that better quality same-sex social interaction in
college was associated with having a larger social network in midlife for men (# =
.29, p < .05), but not for women (# = -.08, ns; see Figure 8). The same pattern of sex-
differences in the relationship between college social activity and midlife social
network size emerged for overall and opposite-sex interaction quality, but both
interactions were marginally significant (both #s = -.18, both ps < .10).
Negative affectivity. College social activity quantity was marginally associated
with reduced negative affectivity in midlife (# = -.17, p < .10). In other words,
participants who had more frequent social interaction in college reported less negative
affectivity in midlife. This effect was predominately driven by marginally lower
midlife negative affect (NA # = -.15, p < .10), and significantly lower midlife
depression (CESD # = -.19, p < .05; SCL90D # = -.19, p < .05); frequency of college
social activity was not associated with midlife anxiety (SCL90A # = -.07, ns) or
perceived stress (PSS # = -.11, ns). The effect of social activity quantity on negative
affectivity was not significant for same-sex social activity (# = -.03, ns) or opposite-
sex social activity (# = -.09, ns). See Table 18 for these results.
Interaction quality was not significantly related to midlife negative affectivity
for overall (# = -.06, ns), same-sex (# = -.08, ns), or opposite-sex (# = -.05, ns)
college social activity.
Positive affectivity. College social activity quantity was positively associated
with midlife positive affectivity (# = .29, p < .01). Participants with more frequent
social activity in college reported greater life satisfaction (# = .24, p < .01), self-
acceptance (# = .26, p < .01), more positive social relationships (# = .30, p < .01),
more positive affect (# = .23, p < .05), and marginally less loneliness (# = -.16, p <
.10) in midlife. The same general pattern was true for same-sex and opposite-sex
college interaction quantity, but the effects were attenuated for the composite social-
activity variable (same-sex # = .14, p < .10; opposite-sex # = .19, p < .05), as well as
the individual positive emotional adjustment measures making up the positive
affectivity composite (see Table 19 for individual regression coefficients).
College social activity quality was marginally positively associated with
midlife positive affectivity (# = .17, p < .10). This effect was predominately driven by
lower midlife loneliness (# = -.24, p <.01), and more positive relationships with
others (# = .23, p < .05); social activity quality was not significantly related to life
satisfaction (# = .05, ns), self-acceptance (# = .10, ns), or positive affect (# = .05, ns).
The pattern was identical for same-sex college social activity quality (positive
affectivity # = .18, p < .05; loneliness # = -.27, p < .01; positive relations with others
# = .29, p < .01), but was not significant for opposite-sex social activity quality. See
Table 19 for regression coefficients.
Self-actualization. Participants with more frequent social activity in college
reported higher levels of self-actualization in midlife (# = .25, p < .01). Time spent
interacting with others in college was positively associated with later personal growth
(# = .24, p < .01), and purpose in life (# = .22, p < .01, and marginally positive
associated with midlife autonomy (# = .16, p < .10). When examining same- and
opposite-sex social activity quantity, the effects on midlife self-actualization are not
significant for either (same-sex quantity # = .08, ns; opposite-sex quantity # = .11,
College interaction quality was not significantly related to later life self-
actualization for overall (# = .03, ns), same- (# = .04, ns), or opposite-sex (# = .01,
ns) social activity. Although the Interaction Quantity X Sex interaction did not reach
significance for the self-actualization component, there was a significant Interaction
Quantity X Sex interaction on one of the measures making up the self-actualization
composite, personal growth (# = -.18, p < .05). Simple slope calculations revealed
that more frequent social activity in college is associated with greater personal growth
for men (# = .44, p < .01), but the effect was not significant for women (# = .08, ns;
see Figure 9). Table 20 contains complete regression coefficients for self-
Daily symptomatology. College social activity was not significantly related to
midlife daily symptomatology (quantity # = .05, ns; quality # = -.05, ns). The same
was true for same-sex social activity (quantity # < .01, ns; quality # = -.05, ns), and
opposite-sex social interaction quality (# = -.09, ns). However, there was a positive
relationship between opposite-sex social interaction quantity in college and midlife
daily symptomatology (# = .20, p < .05). Participants who had more frequent social
activity in college reported experiencing more symptoms in the 4-weeks prior to
completing the current project (# = .19, p < .05). The effect was not significant for
global evaluations of health status (SF-36 # = -.13, ns) or respiratory conditions (# =
.14, ns). See Table 21 for a complete summary of effects
Cardiac health I. Social interaction quantity in college was not significantly
related to the first set of midlife cardiac health outcomes (cardiac health I: cardiac
diagnosis history, systolic and diastolic blood pressure) for overall (quantity # = -.09,
ns), same- (quantity # = -.08, ns) or opposite-sex (quantity # = -.07, ns) social
College social activity quality was, however, associated with enhanced
midlife cardiac health (# = -.21, p < .05). The effect was significant for same-sex
interaction quality (# = -.23, p < .01) and marginally significant for opposite sex
interaction quality (# = -.17, p < .10). Closer inspection revealed that participants who
had more intimate and satisfying interactions in college reported lower systolic blood
pressure in midlife (overall # = -.25, p < .01; same-sex # = -.28, p < .01; opposite-sex
# = -.19, p < .10). The effects for cardiac diagnosis history and diastolic blood
pressure were weaker. See Table 22 for the relevant statistics.
There was a significant Social Activity Quality X Sex interaction on one of
the individual cardiac health measures, cardiac diagnosis history (# = -.21, p < .05).
Calculation of simple slopes revealed that higher quality interactions in college were
associated with fewer midlife cardiac diagnoses for women (# = -.26, p < .05),
whereas the effect was not significant for men (# = .16, ns; see Figure 10). An
identical pattern of sex differences emerged on cardiac diagnosis history for opposite-
sex social interaction quality (# = -.20, p < .05), but the interaction was weaker and
not significant for same-sex college social interaction quality (# = -.15, ns).
Cardiac health II. College social activity was not significantly related to
midlife cardiac health for the second set of cardiac outcomes (cardiac health II:
cholesterol, BMI, and endocrine diagnoses; quantity # = .09, ns; quality # = -.07, ns).
The same was true for same-sex (quantity # = .05, ns; quality # = -.07, ns), and
opposite-sex (quantity # = .02, ns; quality # = -.10, ns) college social activity. See
Table 23 for the regression results.
There was a significant Social Activity Quality X Sex interaction on the
cardiac health II composite (# = -.24, p < .05). As with cardiac diagnosis history,
calculation of simple slopes revealed that better quality social activity in college was
associated with a reduction in cardiac problems for women (# = -.23, p < .05),
whereas the effect was not significant for men (# = .25, ns; see Figure 11). The same
pattern of sex differences in this effect was evident for same- (# = -.22, p < .05) and
opposite-sex (# = -.23, p < .05) college social interaction quality.
Neuroendocrine function. College social activity was not significantly related
to cortisol concentration (AUCg; quantity # = .14, quality # = -.08, ns) or change in
cortisol values over the course of the day (AUCi; quantity # = -.08, quality # = -.04).
The same was true for same- and opposite-sex college social activity; neither quantity
nor quality was a significant predictor of cortisol concentration or cortisol slope. See
Table 24 for regression coefficients.
Summary of longitudinal findings. Table 25 contains a summary of the
overall effects of social activity quantity and quality, sex, and the Quantity X Sex and
Quality X Sex interactions on the nine social well-being, emotional adjustment, and
physical health composite variables, the two cortisol variables, and romantic
intimacy. College social interaction quantity showed the hypothesized effects on three
of four social well-being outcomes, and all three emotional adjustment outcomes, but
no physical health outcomes. College social interaction quality showed the predicted
effects on two of four social well-being outcomes, one of three emotional adjustment
outcomes (though the effect was marginal), and one self-reported physical health
outcome. College interaction quantity was thus a more consistent predictor of midlife
adjustment than college interaction quality.
Controlling midlife social relationship quality. Given that college social
activity was positively and significantly correlated with midlfe social relationship
quality, it might be argued that the effects of college social activity on midlife health
and well-being outcomes are attributable to current social well being. Indeed, there is
some continuity in social activity from college to early adulthood (Reis et al., 1993),
and individuals are likely to reap benefits of immediately available positive social
relationships. However, early adult social activity was expected to exert unique
effects on midlife outcomes. To examine this hypothesis, a set of regression analyses
was conducted to examine the effects of college social activity on midlife emotional
adjustment, physical health, and neuroendocrine function, controlling for midlife
social well-being outcomes.
These analyses revealed the effects of college social interaction quantity on
midlife emotional adjustment to be relatively unchanged (see Table 26). Although the
marginal effect of social activity quantity on midlife negative affectivity (# = -.17, p
< .10) was attenuated when controlling midlife social well-being (#s ranged from -.06
to -.14, depending on which social well-being variable was controlled, all ns), social
activity quantity remained a significant predictor of midlife positive affectivity and
self-actualization when controlling romantic relationship quality (positive affectivity
# = .21, self-actualization # = .20, both p’s < .05), friendship quality (positive
affectivity # = .17, self-actualization # = .18, both p’s < .05), and social network
(positive affectivity # = .21, self-actualization # = .21, both p’s < .05).
The effects of college social activity quality on the first midlife cardiac health
composite were also relatively robust. Social activity quality continued to predict
cardiac health I when controlling midlife romantic relationship quality (# = -.26, p <
.05), friendship quality (# = -.18, p < .10), and social network (# = -.21, p < .05).
In sum, controlling midlife social relationship quality seemed to have little
impact on the other longitudinal findings, suggesting that the predictive effects cannot
simply be reduced to current social circumstances, or long-term continuity in social
Controlling personality. Participants’ perceptions of their social worlds, and
thus their self-reports of social activity, social relationship quality, and emotional
adjustment may be subject to perceptual biases associated with certain personality
traits. In an attempt to discern whether personality factors were responsible for the
obtained effects, analyses were conducted in which the two most social interaction
relevant personality characteristics were controlled: neuroticism and extraversion.
Results are presented in Table 27.
The first set of analyses attempted to replicate the longitudinal findings while
controlling neuroticism, a dispositional bias toward perceiving and experiencing the
world negatively (John & Srivastava, 1999). The longitudinal effects of social
interaction quantity and quality on social, emotional, and health outcomes remained
virtually unchanged with neuroticism controlled. A comparison of the neuroticism-
controlled effects presented in the left side of Table 27 to the uncontrolled effects
presented in Table 25 revealed that the only substantive changes were that the two
marginal effects (of college interaction quantity on negative affectivity, and college
interaction quality on positive affectivity) became significant (respective #s = -.15
and .16, both ps <.05) with neuroticism controlled.
The second set of analyses controlled extraversion, a personality trait
associated with a high level of sociability and positive emotionality (John &
Srivastava, 1999). Controlling extraversion weakened some of the effects of college
interaction quantity (though not quality) on midlife social relationship quality (see
right side of Table 27). The effects of college interaction quantity on midlife romantic
intimacy (# = .14), friendship quality (# = .06), negative affectivity (# = -.13), and
self-actualization (# = .10) were reduced to non-significance (all ps > .05. The effect
of interaction quantity on positive affectivity was attenuated as well, but remained
significant (# = .17, p < .01). The effects of college interaction quality, however, were
more robust when extraversion was controlled. College social interaction quality
continued to predict romantic intimacy (# = .21, p < .05), friendship quality (# = .24,
p < .01), and the first cardiovascular health variable (# = -.23, p < .01). One college
social interaction quality effect was attenuated, the effect on midlife positive
affectivity (# = .12, ns).
Where personality variables did impact the findings of this research, it is
noteworthy that the less subjective college social activity variable (interaction
quantity), which is expected to be least vulnerable to self-report biases associated
with personality traits, was attenuated when extraversion was controlled.
Nevertheless, these results must be interpreted with caution because personality was
assessed currently, and not in college. Although personality is said to remain stable
across time, it is unclear how socialization in early adulthood may impact later self-
reports of personality. An assessment of neuroticism and extraversion in college
would allow more confidence in these results, a point I will address further in the
Chapter 4: Discussion
This research explored the interconnections among social, emotional, and
physical health in a sample of midlife (age 48-52) adults. Concurrent relationships
between midlife health and well-being were investigated as well as longitudinal
effects of early adult (age 18-22) social engagement on the same midlife outcomes. A
summary and discussion of concurrent findings will be followed by a summary and
discussion of longitudinal results.
Summary of Concurrent Findings
The concurrent results of this research are consistent with some widely
accepted, and empirically supported basic assumptions about the interplay between
social relationship quality and emotional well-being. Midlife adults who were better
socially adjusted (i.e., those who reported better romantic and friend relationships,
and larger and more diverse social networks) also reported being better emotionally
adjusted (e.g., they experienced less negative affect, depression, anxiety, stress, and
loneliness, and more positive affect, life satisfaction, positive relationships, self-
acceptance, purpose, growth, and autonomy). Fifteen of 16 composite variables were
significantly correlated, and one correlation was marginally significant, all in the
expected direction (shown in the top two panels of Table 11). Among the individual
variables making up the composites, 57 of 66 correlations were significant, and three
were marginal (shown in the left two panels of Table 12). This is consistent with prior
theory and research indicating that social and emotional adjustment are reciprocally
related (Arnstein, 1984; Bowlby, 1969, 1979; Harker & Keltner, 2001; Hartup &
Stevens, 1997; Reis, Collins, & Berscheid, 2000; Rook, 1984). However, whereas
many studies show one facet of social life to be related to one or two emotional
adjustment outcomes, the current research incorporated a variety of measures to
capture multiple dimensions of social and emotional well-being, couching individual
associations between social and emotional life in a broader context of positive
As for self-reported physical health, only one significant correlation with
midlife social well-being emerged: People who felt more closely connected with their
core group of friends also reported better cardiac health -- that they were less likely to
have been previously diagnosed with cardiovascular conditions. Although this single
result might be due to chance, the finding fits well with existing research showing
that good social ties enhance cardiac health. It is consistent with Uchino and
colleagues’ work showing that a greater number of positive and supportive
relationships reduce cardiovascular reactivity to stress (Uchino et al., 2001; Uchino et
al., 1992), and also dovetails with other work showing that supportive social ties
enhance recovery and prognosis following acute cardiovascular problems such as
congestive heart failure (Coyne et al. 2001), heart attack (Helgeson, 1991), and
coronary artery bypass surgery (King et al, 1993). However, cardiac diagnosis history
was the only measure of physical health that was significantly correlated with any
measure of social well-being, and it was correlated with friendship quality, not
romantic relationship quality. Several of the studies mentioned above have been
conducted with married couples, and the effects have been shown for spousal support.
Only Uchino’s work looks at cardiac health in the context of other types of
relationships. Why friendship quality, but not romantic relationship quality, predicts
cardiac health in this research remains unclear. It is noteworthy that this particular
outcome was relatively less subjective as far as self-report measures go (in that
participants were asked to indicate how many of several cardiac disease or disorders
had been previously diagnosed by a doctor). Given the outside source of diagnosis,
this measure of health may be less susceptible to the biases associated with memory
or personality traits like neuroticism discussed previously.
Another noteworthy finding that emerged from the concurrent analysis was
the correlation between positive emotional adjustment and cardiac health (again, a
reduction in cardiovascular condition diagnoses). Some evidence has suggested that
the health benefits associated with positive and supportive ties may reflect a social
life enriched with positive emotional experiences. Other research has found short-
term and long-term health benefits of emotional experience and expression (Danner,
Snowden, & Friesen, 2001; Fredrickson, 2000; Fredrickson & Levenson, 1998;
Harker & Keltner, 2001; Mendes et al., 2003), and theorized that social life is the
mechanism. Broadened social resources (Fredrickson, 2000) and more frequent
engagement in high quality social interaction (Keltner, 2003) are two explanations.
This chicken-and-egg problem yet remains unsolved. Although mediation analysis is
statistically possible (and reveals that friendship quality, rather positive emotional
adjustment, remains the significant predictor of cardiac health when both variables
are entered into a regression equation), the lack of temporal difference in assessment
of the two variables prevents an unequivocal causal statement. In all likelihood,
others gravitate toward happy people who express positivity because they anticipate
rewarding social interactions with them (Byrne & Clore, 1970). Disclosure of positive
emotion is well received and reinforced by interaction partners, fostering further
positive emotional experience, and creating a social and emotional upward spiral.
Summary of Longitudinal Findings
The results of this research partially supported the longitudinal hypotheses.
The first hypothesis concerned midlife social well-being, and was largely supported;
social activity in college was generally associated with better midlife social outcomes.
High quantity and high quality college social activity was associated with better
friendships -- that is, friendships characterized by greater perceived support, closeness
and intimacy with a core group of confidants -- in midlife. Individuals who reported
more frequent and better quality social encounters in college also reported more
intimate relationships with their romantic partners as adults. Moreover, frequent
social interaction in college was associated with participant reports of larger and more
diverse social network in midlife, though the quality of college social activity was
unrelated to social network size in midlife. Thus, frequent socializers in college were
well connected to sizeable, diverse networks in middle adulthood, a pattern that
presumably remained stable throughout adulthood.
The emotional adjustment hypothesis was also largely supported: Social
activity in college was generally associated with participant reports of enhanced
emotional adjustment in midlife. Frequent social activity in college was associated
with higher levels of midlife hedonic and eudaimonic well-being (less negative affect,
depression, and loneliness, and more positive affect, life satisfaction, self-acceptance,
positive relationships, personal growth, purpose, and autonomy). High quality (i.e.,
intimate and satisfying) social activity in college was also associated with certain
positive emotional adjustment outcomes in midlife, particularly those relating to
positive social relationships (i.e., the absence of loneliness and the presence of
The emotional adjustment findings are consistent with Hartup and Stevens's
(1997) theorizing that positive social engagement throughout life facilitates
adaptation and contributes to well-being. Their review indicated that a major benefit
of having friends in early life is the enhancement of feelings of self-worth. The
current research suggests that this carries through life because those who were more
socially embedded in college seemed to accept, value, and feel more positively about
themselves in midlife.
By and large, the quantity of social activity during college was more strongly
related to midlife psychological well-being than was the quality of social activity
during college (which was also generally true for social well-being outcomes).
College interaction quantity effects were relatively robust across the emotional
adjustment composites and individual measures, with the exception of two midlife
outcomes: anxiety and perceived stress. Perhaps fluctuations in these two variables
are more vulnerable to small environmental variations (e.g., a particularly stressful
day, week, or month may be more likely to alter perceptions, and thus reports of
anxiety and stress) than are the other variables included in the emotional adjustment
composites (e.g., depression, life satisfaction, self-acceptance, positive and negative
affect, loneliness, and positive relationships with others). Effects were less consistent
for college interaction quality, which seemed to best predict the two midlife measures
that were theoretically most closely associated with interaction quality: loneliness,
and positive relationships with others. I will address this in greater detail below, when
I discuss the relative effects of structural and functional support.
The results for the physical health hypothesis were generally weaker.
However, there was some evidence that midlife cardiac health was enhanced among
individuals who reported having high quality social activity in college – one of two
cardiac health composites was predicted by college social interaction quality, and the
effect was driven by a reduction in self-reported systolic blood pressure. Consistent
with this finding, there was also a negative trend between college social interaction
quality and cardiac diagnosis history, such that people with more intimate and
satisfying social lives in college were less likely to report having been diagnosed with
high cholesterol, high blood pressure, coronary heart disease, angina, heart attack,
stroke, or any other heart condition by the time they had reached midlife. This trend,
however, did not reach significance.
The effect sizes obtained in the current study were modest. The proportion of
variance explained (R2) by college social activity ranged from .03 for the smallest
marginal effect (on negative affectivity) to .12 for the largest significant effect (on
friendship quality). However, considering these effects were obtained over 30 years,
small effect sizes are not surprising, and the consistency of the results within each
category of outcomes lends further credibility to the authenticity of these findings.
Unsupported Longitudinal Hypotheses
Notwithstanding the effects that were uncovered, several noteworthy non-
significant results emerged as well, particularly for romantic relationship and physical
Romantic relationship quality. Although college social activity was related
to midlife romantic intimacy, it did not significantly predict romantic relationship
status, or quality (e.g., closeness, support, responsiveness) for those who were
involved in committed, monogamous relationships. It is interesting that romantic
intimacy was predicted by college social life, whereas the other measures of romantic
relationship quality were not. Intimacy was positively and significantly correlated
with each of the other measures of romantic relationship quality (rs ranged from .37
to .46, all ps < .001), but loaded with friendship quality variables in the principal
components analysis. The elements of intimacy that are predicted by early adult
socialization apparently differ from the elements that intimacy shares with feeling
close to, satisfied with, and supported and responded to by a romantic partner. This
may be partly explained by examining the different facets of intimacy assessed by the
PAIR. Some subscales addressed the extent to which romantic partners engage, as a
couple, in activities and relationships external to their primary relationship, such as
spending time with other couples, sharing interests, and mutually enjoying fun
recreational activities (e.g., social and recreational intimacy). Other subscales (e.g.,
emotional and intellectual intimacy) are more similar to the other romantic
relationship specific measures included in this study (perceived support, closeness,
satisfaction, and responsiveness), which more specifically focused on participants’
feelings about the romantic partner, and what the romantic partner provides to the
participant within the dyad. In an ancillary analysis, a closer examination of the
subscales revealed college social activity was more strongly related to the social and
recreational intimacy subscales of the PAIR (both quantity #s > .17, both ps < .07)
than to the emotional or intellectual intimacy subscales of the PAIR (both quantity #s
< .15, ns). Thus, socialization in college predicted a better social life in the context of
later romantic relationships (what may be characterized as an interpersonal benefit),
but was not related to specific emotional benefits derived solely from the romantic
partner (perhaps characterized as an intrapersonal benefit).
It is somewhat surprising that midlife romantic relationship quality was not
predicted by early adult social life. Childhood friendships are said to be the precursors
to adolescent romantic life (e.g., Hartup & Stevens, 1997), thus it would seem to
follow that early adult social engagement would foster later life romantic ties. Hartup
and Stevens (1997) argued that same-sex relationships in childhood and adolescence
set the stage for fulfilling intimacy needs in early adulthood. Once established, they
may play a key role in later romantic relationship development by preparing young
adults for the opposite-sex socializing that facilitates the pivotal relationship-relevant
developmental transitions common to early adulthood. These transitions include
finding a romantic partner, getting married, and starting a family, as described by
Erikson (1959), Levinson (1978), White (1966), and others (e.g., Neugarten et al.,
1965). Sex-specific social activity was examined to see if it would account for the
lack of longitudinal effects on romantic relationship quality, but there was little
difference in midlife romantic relationship quality as a function of same-sex versus
opposite-sex social engagement in college. In fact, neither same-sex nor opposite-sex
social activity in college predicted midlife romantic relationship quality. Midlife
romantic intimacy, however, was predicted by same-sex and opposite-sex college
social activity quality (but not quantity). This is similar to the sex-specific effects for
midlife friendship quality outcomes, where both same- and opposite-sex social
activity show similar patterns of prediction – college social activity quality, but not
quantity, was associated with better midlife friendships.
Just as same-sex friends play an important role in bolstering well-being and
feelings of self-worth through adolescence (Hartup & Stevens, 1997), the current
research shows they continue to do so into midlife. However, the results of this study
suggest a somewhat different pattern than Hartup and Stevens discussed. Although
same-sex social engagement was still important for many aspects of adaptation and
adjustment, in the current sample it was not instrumental in later romantic life (the
longitudinal effects of same-sex social activity on midlife romantic relationship
quality were not significant). It also appeared that while same-sex relationships
remain important in early adulthood (e.g., same-sex social activity in college
longitudinally predicted enhanced midlife friendship quality and positive affectivity),
as Hartup and Stevens suggest, opposite-sex relationships gain importance (e.g.,
longitudinal effects of opposite-sex college social activity on midlife friendship
quality and positive affectivity were also significant). Although midlife romantic
relationship quality was not predicted by same- or opposite-sex social activity, other
important outcomes were. Apparently socializing with opposite-sex peers in early
adulthood becomes increasingly important to later-life adjustment, notwithstanding
the development of a romantic relationship, or the quality of that relationship.
Physical health. Several studies have shown evidence for subjective and
objective physical health gains as a function of social integration (Berkman & Syme,
1979; Cacioppo et al., 2000; Cohen et al., 1997; House et al., 1988) and positive
social interaction (Baker et al., 2000; Berkman, 1995; Carels et al., 1998; Coyne et
al., 2001; Gump et al., 2001; King et al., 1993; Robles et al., 2006; Uchino et al.,
2001). However, the current research generally failed to replicate these effects. For
the most part, subjective physical health outcomes and neuroendocrine function were
unrelated to concurrent measures of social well-being or emotional adjustment, or
college social activity.
Although one of the cardiac health composites was predicted by both college
interaction quality, and midlife friendship quality, daily healthfulness outcomes (e.g.,
physical functioning, pain, participants’ global self-perceptions of healthfulness, daily
experience of specific symptoms, or respiratory problems including asthma and
sinusitis) and other measure of cardiac health (e.g., cholesterol, BMI, and endocrine
system disorders such as diabetes) were unrelated to college social activity variables,
or midlife social and emotional adjustment variables. Concurrently three of eighteen
correlations (16.7%) with subjective physical health outcomes were significant, and
longitudinally, one of six effects (16.7%) of college social activity on self-reported
physical health was significant. Moreover, despite several studies indicating that
social relationships and social activity are related to HPA axis activity and
neuroendocrine function (Cacioppo et al., 2002; Kiecolt-Glaser et al., 1998; Malarkey
et al., 1994; Robles et al., 2006), neither the midlife measures of social well-being nor
the college measures of social engagement were significant predictors of cortisol
concentration or diurnal rhythm. The lack of findings begs explanation, and (apart
from the straightforward explanation that they are simply unrelated) two possible
alternative explanations come to mind: measurement issues and health status.
Measurement concerns. The measurement of social life in this study,
particularly in college, was quite different from measures typically employed in
research of this nature. In assessing social activity with the RIR this research captured
facets of social life that are typically unrepresented. Social life is typically measured
in terms of social integration (e.g., Berkman & Syme, 1979; Cohen et al., 1997;
House et al., 1988), or global perceptions of relationship quality in a general sense
(e.g., Cacioppo et al., 2000; Uchino et al., 2001), or with a specific target (e.g., Baker
et al., 2000; Coyne et al., 2001; Gump et al., 2001; King et al., 1993). This is among
the first research to use such rich and in-depth measures of social activity to represent
social life, particularly in a longitudinal study. The RIR captures social activity across
diverse relationships, ranging from superficial to exclusive, from frequent interaction
partners to chance encounters, and combines across all types of relationships and
interactions to provide what is arguably a relatively more objective picture of social
life than that captured by more global measures. Undoubtedly, all kinds of measures
are informative and useful, however, the different results seem to possibly reflect a
pattern in which more global measures appear to better relate to health. Perhaps the
ways in which individuals subjectively weave together the details of their social lives
to build a coherent story is partly responsible for these effects.
Whereas the measurement of social life may have been relatively more
objective than is customary, the current measurement of health was relatively more
subjective than is typical with this type or research. As reviewed above, health effects
have been observed for a large range of outcomes, all of which were objectively
evaluated by a blood test, a biomarker, or medical professional. The assessment of
self-reported health in this research may have been influenced by biases associated
with self-report measures, as discussed previously. Also, as noted above, the least
subjective measure of health (prior diagnoses) was one of the only health outcomes to
show an effect.
Neuroendocrine function was the only truly objective biological measure
included in this study, and may have been affected by the time frame of assessment
and the level of specificity with which it was measured. Many studies showing an
association between social variables and HPA axis activity measure both variables in
the same session, looking, for example, at how cortisol levels fluctuate as a function
of interacting with a specific significant other, or looking at how the presence of a
meaningful other alters HPA axis reactivity to an outside stimulus. Given the
dispersion of participants across the United States and abroad, it was not possible to
collect this data in the lab, where the simultaneity of the self-report and saliva data
could have been ensured. This design prioritized the flexible accommodation of
participant’s schedules in order to maximize the amount of data obtained by allowing
take-home materials to be completed on a self-paced timetable. Thus, the time frame
between saliva sampling and survey completion varied among participants. To be
sure, I explicitly made participants aware of the importance of adhering to the saliva-
sampling schedule and completing the surveys in a timely manner. An electronic time
stamp on survey submissions made compliance transparent, but adherence to saliva
sampling procedures could not be verified by anything other than the participant’s
report. Some participants completed all tasks on the same day, whereas others spread
them out over several weeks. More than 75% of participants completed all tasks
within a three-week time frame. Nevertheless, concurrent effects involving
neuroendocrine function were not obtained.
Second, perhaps the slice of HPA axis activity that was captured to represent
neuroendocrine function was too narrow. Cortisol was sampled on one day, and it is
unclear whether that day was adequately representative of a participant’s typical HPA
axis activity. Although participants were asked to select a day that they expected to be
typical, the day could have turned out to be unexpectedly stressful or relaxed.
Perceived stress was not assessed on the sampling day, and thus could not be
Health status. Attempting to reconcile the non-significant findings in this
study with prior longitudinal findings (Berkman & Syme, 1979; Holt-Lundstad,
Smith, & Layton, 2010; House, Landis, & Umberson, 1988) brings to mind a second
plausible explanation for the lack of health effects in this research. The current
sample, just approaching their early fifties, is barely on the precipice of a time frame
when health begins to deteriorate. As it stands, there was a relatively low incidence of
illness in the sample. Only 16.5% of participants had been diagnosed with respiratory
illness, less than 40% had been diagnosed with any kind of cardiac issue, and only
21% had been diagnosed with any kind of endocrine system problem. Fewer than
10% of the sample had dealt with cancer and only one participant had kidney or liver
problems. Ninety percent of the sample reported daily symptom (CHIPS) scores of
less than .50 on a 0-4 scale and more than 70% of participants earned a score of 85 or
better (out of 100) on the physical health subcales of the SF-36. Taken together, it
appears that the sample was generally quite healthy. Prior research has examined
samples of a wider age range, which include a larger number of older participants.
For studies in which the population was already suffering a serious health
complication, participants were older, on average, than the current sample, and
contained a greater number of older participants (e.g., average age = 53 for Coyne et
al., 2001, with participants up to 78 years of age; average age = 60 for King et al.,
1993, with participants up to 80 years of age). Where mortality risk was the key
outcome of interest, it was most common among the highest age bracket (e.g. 60-69
years of age; Berkman & Syme, 1979). At the time of assessment nine participants
from the original pool of 222 had died (college data for 8 was available). Although an
n of eight is insufficient for comparison, an analysis was undertaken comparing the
college social activity of the eight deceased participants with the college data from
the 163 of 171 participants who were confirmed to be alive at the time of current
assessment (college data for 8 living participants were missing).7 No differences in
college social activity were uncovered.
As discussed previously, the age range of the participants at the time of
current assessment seemed ideal, given the general tendency for onset of several
serious health problems at this juncture. However, in this healthy sample, health
problems were relatively uncommon and showed little variance. Socioeconomic
status (SES) may provide one plausible explanation for the lack of health findings.
Poor socioeconomic standing is associated with greater susceptibility to illness and
disease (Adler et al., 2000; Williams & Collins, 1995), yet the current sample was
relatively comfortable, well educated, well employed, and financially secure. Over
98% were college graduates and more than 70% of the sample reported having an
advanced degree. Seventy percent indicated that it was “not at all difficult” to meet
expenses, just over 90% of the sample saw themselves at step seven or higher on
Adler et al.’s (2000) 10-rung ladder of subjective socioeconomic status, and fewer
than 15% indicated that there was not enough money to meet their family's needs.
Thus, given the relative lack of variability in subjective SES, it comes as no surprise
that controlling for SES left the findings of this research virtually unaltered. The
sample's relatively high socioeconomic standing may have served as a health-
protective factor, delaying the onset of vulnerability to health problems. Seeman
(1996) reported that evidence for mortality and mental health outcomes was stronger
and more consistent than evidence for other physical health outcomes, especially
when looking at reasonably healthy people. Although it may be premature to examine
mortality in this relatively young sample, mental health (i.e., emotional adjustment)
outcomes showed some of the strongest effects in this research. It is particularly
interesting to note that the cardiac diagnosis history variable, one of the only physical
health variables for which effects of early adult social activity were found, was the
most frequently endorsed health issue. It was nearly twice as common as the next
closest health issue (endocrine related problems, such as diabetes or high blood
sugar). It would be informative to track this sample over time, and observe whether
health disparities related to early life social activity emerge as participants continue to
age and health concerns become increasingly prevalent.
In sum, the lack of significant health results might be a function of
measurement or procedural problems, or might reflect a true lack of association
between early adult social activity and midlife health (an explanation deemed less
likely given extant published research indicating otherwise), or perhaps this sample
was too young and healthy at the time of assessment to uncover health effects.
His, Hers, or Theirs: Sex Differences
Sex differences (or lack thereof) also warrant discussion. Several mean
differences between men and women were uncovered on midlife outcomes. Women
reported better friendships than men did, and scored higher on positive affectivity
outcomes, whereas men reported more cardiac health problems than women did.
These mean differences are not surprising; women tend to have more emotional
closeness, and share greater intimacy with friends than do men (e.g., Caldwell &
Peplau, 1982; Hartup, 1993; Wheeler & Nezlek, 1977), and it is widely known that
men more frequently suffer from heart-related problems than do women (e.g.,
Nikiforov & Mamaev, 1998). However, and perhaps more interesting, men and
women were overwhelmingly similar in the concurrent and longitudinal associations
among social, emotional, and health variables that were uncovered. Despite mean
differences in some of the variables that were measured, the pattern of associations
among the variables was less likely to show sex differences. Of 55 concurrent
correlations among social, emotional, and physical health outcomes, the results of a
comparison of correlation coefficients revealed men and women differed significantly
in the magnitude of only one (1.8%) effect (the correlation between the two cardiac
health composites was significant and strong for women, but much weaker for men;
see Table 14). Longitudinally, of 54 possible overall, same-, and opposite-sex
quantity and quality effects on 18 composite health and well being outcomes, only 3
(5.6%) were significantly moderated by sex (the effect of overall, same- and opposite-
sex social interaction quality on the second cardiac health composite).
Although research has shown that social factors provide protective benefits
for men and women (e.g. Holt-Lundstad, Smith, & Layton, 2010; Kiecolt-Glaser &
Newton, 2001), the literature diverges with regard to whether the benefits of social
support and social integration apply equally to the two sexes. Some authors have
argued that social relationships are more important to women's well being (e.g.,
Glenn, 1975; Taylor et al. 2000; Wood, Rhodes, & Whelan, 1989). Women, who, for
a variety of reasons including socialization and evolution (Eagly & Wood, 1999),
place greater value on social relationships, are more likely to cope with stress by
affiliating (as opposed to the standard male fight-or-flight response; Taylor et al.,
2000), and have friendships that are characterized by higher levels of self-disclosure
and intimacy (Caldwell & Peplau, 1982; Hartup, 1993). Thus, social ties are
presumed to be more instrumental in shaping women's health and well-being.
However, other authors have suggested that men reap greater benefits from good
social ties, particularly a good marriage (e.g., Diener et al., 1999). Although married
persons generally tend to be physically and mentally better off than their unmarried
counterparts, sex discrepancies still emerge in some studies. For example, the
reduction in mortality risk associated with marriage was, in one review, estimated to
be five times greater for men than for women (Kiecolt-Glaser & Newton, 2001). Two
popular explanations for this divergence are 1) that wives are more likely to exert
positive influence by taking an active role in the management their husbands' health,
and 2) whereas women have a variety of social connections beyond marriage to
provide health and well-being protective benefits, men derive the majority of such
benefits from the marital relationship (Kiecolt-Glaser & Newton, 2001; Umberson,
2002). Despite the controversy over which sex gains more from an enriched social
life, the current data suggest that both men and women derive equivalent short-term
and long-term benefits from social connection (particularly good friendships) and that
those benefits come in the form of enhanced long-term social and emotional
adjustment. Very few of the longitudinal effects were moderated by sex, indicating
that benefits garnered from early adult social engagement did not vary significantly
by sex. The few effects that were moderated by sex did not show one sex to
experience a clear and consistent advantage over the other.
These findings do not resolve the sex-difference debate, as physical health
gains were not predicted by social life for either sex, by either romantic ties or
friendships. Potential explanations for the lack of physical health results were
discussed above. Mortality, the variable that has shown the biggest disparity for the
two sexes in some research (e.g., Kiecolt-Glaser & Newton, 2001), could not be
tested reliably as an outcome because of the small number of deceased participants.
However, others like Holt-Lundstad, Smith, & Layton (2010) find sex similarities,
and in the current research men were just as likely as women to enjoy emotional and
social benefits of being socially integrated in early adulthood, and having good
quality social connections in midlife, contrary to some prior research and theory
(Diener et al., 1999; Kiecolt-Glaser & Newton, 2001).
Assessing Support: Form Versus Function
The relative merits of structural versus functional support were previously
discussed, and although empirical evidence suggests that there are benefits of both
kinds of support (e.g., Berkman, 1995; Berkman & Syme, 1979; Cacioppo et al.,
2000; Cacioppo et al., 2002; Cohen et al., 1985; House et al., 1988; Uchino et al.,
1996), rarely are the two examined simultaneously. The current research set out to
examine differences between structural (social integration/quantity) and functional
(perceived support/quality) indicators of social support. Generally speaking, although
both quantity and quality of college social activity were influential in longitudinal
analyses, interaction quantity was a stronger and more consistent predictor of long-
term health and well-being than quality of college social activity. Although simply
having social connections has been shown to provide benefits, the nature of those
relationships has also been deemed important. For example, several findings indicate
that ambivalent or hostile social ties are detrimental to health and well-being (Kiecolt-
Glaser & Newton, 2001; Uchino et al., 2001).
Many prior findings about relationship quality have been obtained in studies
with a shorter time span and with greater specificity. That is, concurrent studies often
examine the correlation between relationship quality with one specific target (e.g. a
spouse, or romantic partner) and a more narrowly focused health outcome (e.g.,
cardiovascular stress reactivity). Taking a broader perspective, as this research did, it
becomes clearer why interaction quantity may have yielded stronger results. The time
frame in which the social interaction data were collected represents a time when
young adults are exploring the social world on their own for the first time, outside the
safety net of their family environment. The time period is developmentally critical; it
is a point at which one has the potential for a great deal of growth. More frequent
interaction at this stage provides more opportunities to learn about oneself, about
others, and about how to navigate the ever-changing social world. Frequent social
contact may teach lasting lessons to be taken forward through life, preparing young
adults to adapt to the challenges that lie before them. This preparation develops self-
knowledge about one’s strengths and weaknesses, arming individuals with coping
skills, and fostering feelings of efficacy when facing obstacles down the road.
Acquiring this knowledge, skill, and ability through high quality social interactions is
presumably even more beneficial. However at this critical stage of social
development in early adulthood, frequent social interaction, which may also imply
many social outlets for learning about the self and the world, appears to be relatively
This explanation is consistent with Baltes and Baltes’s (1990) model of
successful lifespan development, which involves selectively focusing on
developmentally relevant goals that provide the most positive gain, optimizing the
attainment of those goals by devoting resources to them, and compensating for any
loss in the ability to achieve those goals (see also Baltes, 1997). Although the
selective optimization with compensation (SOC) model (Baltes & Baltes, 1990) is
applied to multiple domains of development, whereby certain domains of functioning
may take priority at any given life phase, Carstensen and colleagues’s theory of
socioemotional selectivity (Carsentsen, 1995; Carstensen, Isaacowitz, & Charles,
1999) has zeroed in on the application of the SOC model to the social domain. Their
research has found that as perceived time grows short, social goals shift from
knowledge acquisition to emotion regulation. Time perspective is unavoidably tied to
age (although not completely dependent upon it - research has also shown a shift
toward emotion regulation goals among younger groups who perceived time to be
limited; Carstensen & Fredrickson, 1998). When participants in this research were in
college, time would likely have been perceived as open ended to the overwhelming
majority, and knowledge acquisition social goals would have reigned supreme.
Frequent social interaction would (and, as evidenced here, did) allow them to
successfully fulfill those goals in the service of long-term adaptive development.
However as time wore on and became conspicuously more limited, it is likely that the
social goals of these participants became ever-more geared toward emotion
regulation, and social interaction quality would have become increasingly critical to
long-term health and well being.
In fact, the quality of the contemporaneous midlife social environment was
beneficial to emotional well-being, and even some indicators of health. However,
analyses looking at the unique effects of early adult social activity on midlife health
and well-being outcomes, while controlling for midlife social well-being, showed that
the effects of early adult social activity on midlife outcomes remained largely
unchanged when controlling midlife social relationship quality. Thus, there is
something unique and important about the early adult social world, particularly the
frequency of socializing, that remains influential in later life.
Longitudinal effects spanning 30 years, such as these, beg the question of
mechanism. In all likelihood, it is probably a combination of factors that contribute to
consistency over time. Temperament, which is partly biologically based (Goldsmith
et al., 1987), likely influenced early social development (long prior to the time period
investigated here), giving rise to scripts that would guide socialization throughout life
(Bowlby, 1969; Sullivan, 1953). These early-developed scripts would have shaped
social interactions throughout adolescence and into adulthood. Socially secure
individuals (i.e., those with positive models of self and other) would have been prone
to approach the social world with a positive and optimistic outlook that facilitated
rewarding social interactions, reinforcing the individual’s views and aiding in the
development of social skills in a self-fulfilling prophecy. For individuals who
followed such a trajectory, the new social milieu encountered in college was yet
another opportunity to apply their skills in achieving the affiliation goals that became
relevant at this developmental stage (Erikson, 1959). Thus, although personality
variables may have had an impact on the socialization processes that took place, the
social experiences in college were necessary for the accumulation of long-term
benefits, which is why these effects cannot be reduced to temperament (or
This general framework suggests a probable normative pattern, and those who
follow it are expected to regularly engage in many of the previously reviewed
relationship processes that have been shown to provide short-term benefits in
controlled laboratory settings. For example, the reduced cardiovascular reactivity to
laboratory-induced stress associated with supportive partners (Uchino et al., 1996),
and the efficient HPA-axis recovery pattern associated with positive behaviors during
a laboratory conflict discussion (Robles et al., 2006) capture a snapshot of the fleeting
processes that are expected to recur on a larger scale in everyday life and to have a
cumulative positive impact on a recipient’s health and well-being. In addition,
regularly engaging with relationship partners in activities that produce positive affect
is also expected to contribute to long-term health and well-being (Lyubomirsky et al.,
2005) in much the same way. Thus, social relationships provide the vehicle by which
various specific health and well-being bolstering mechanisms are able to manifest.
Several limitations to the current research warrant note. First, the sample was
relatively homogenous. Participants came from the same university, and the social
life they experienced in that setting may not generalize to social life experienced by
students at other colleges and universities, young adults who lived at home and
commuted to college, or those who did not attend college at all. Persons who did not
have the residential college experience may not have the same kinds of social
experiences that come with living in a self-contained residential environment (e.g,
planned activities, meals in a shared facility) for four years of early adulthood, when
the primary developmental social goal is establishing intimacy and sharing one’s life
with others (Erikson, 1959). It would be important to determine whether similar
findings would be obtained among non-college students to understand if the
transitions that occur at this time are developmental transitions associated with a
phase of life, or rather a function of the unique social circumstances associated with
attending a residential college.
Second, the current sample was composed of relatively happy, well-adjusted,
well-educated, financially secure, mostly white, midlife adults, who lived in
residential halls during college, which maximized opportunities for social contact.
The findings uncovered in this research may not generalize to more diverse samples.
However, if anything, this sample’s relatively high level of socioeconomic status, low
level of social isolation, and high level of emotional adjustment may have worked
against the longitudinal hypotheses. As discussed above, the socioeconomic
advantage enjoyed by this group of adults may have protected them from some of the
health pitfalls common to this age group. If so, early-adult social activity might be a
stronger predictor of health problems in a relatively more impoverished sample.
Those who are more susceptible to poor health may also have more to gain from
frequent, supportive socializing. Examining these associations with a wider range of
socioeconomic classes would illuminate this question. Moreover, although there was
variability in college social activity, social integration may have been skewed toward
the high end in this sample because the social environment of a college residential
living facility afforded participants constant, built-in opportunities for socializing.
Socializing is central to what college students do on a day-to-day basis, and having
been surrounded by roommates, hall-mates, and classmates around the clock fostered
participants’ ability to be socially integrated if they chose to take advantage of their
environment. Examining less socially embedded people may improve our ability to
detect health disparities as a function of social activity.
Third, this research was limited to a single birth cohort. These participants,
born in the mid 1950s, lived through a unique set of social circumstances. Growing
up in a time when social ideals were shifting from the traditional wife-as-homemaker
and husband-as-breadwinner model, through periods of free love and women’s
liberation, presented varied models for lifestyle choices about work, marriage, social
relations, and parenting. The first wave of research with this sample, conducted in the
late 1970s, may have caught participants just on the tail end of a time when the
transition from college to the work world was associated with marriage and starting a
family. Marital status data for the U.S. (as reported by the United Nations, 2009)
revealed that in 1980 (when the last of this cohort graduated college, and the youngest
was about 21 years old), 41.4% of women and 27.9% of men in the U.S. aged 20-24
were married. Twenty years later, in 2000, those figures had dropped to 22.5% for
women and 12.9% for men. Gender roles and social scripts had become ever more
loose and varied. The transitions made by this sample may not have the same
meaning for today's college students, who often put off goals of marriage and family
in favor of developing a career. Women in the current study were only just beginning
to make such a choice more commonly. Women born in previous cohorts may have
attended college with the goal of attracting a husband looming larger than educational
aspirations, or may never attended college at all, and missed out on this cohort’s
social experience altogether. It would be informative to replicate this research with a
new cohort of college students, following them into midlife to see what impact, if
any, shifting social scripts and standards may have on the findings uncovered here.
A fourth limitation concerns the nature of the measures. Notwithstanding the
previously discussed benefits of experience sampling over other kinds of self-report
measures, the RIR remains a somewhat subjective self-report measure. Although
retrospection bias may be less of a concern because reports are completed online, the
tendency for individuals to pessimistically peer out at the world from under a dark
rain cloud and devalue the quality of their social interactions (or, conversely, to see
the world through rose-colored glasses) may still have resulted in reporting biases. In
an attempt to unravel these kinds of personality effects, a series of analyses were
undertaken in which neuroticism and extraversion were controlled. The effects of
college social activity (quantity and quality) were virtually identical when
neuroticism was controlled. However, when extraversion was controlled, the effects
of college social interaction quantity (but not the effects of quality) were weakened. It
is noteworthy that the quantity variable, which was altered by extraversion, should
have been less susceptible to global biases because of its relatively more objective
nature (the length, and number of interactions). On the other hand, the more
subjective (and thus more vulnerable to global reporting biases) quality variable
(consisting of reports of interaction intimacy and satisfaction), was not altered when
extraversion was controlled, suggesting that global positivity reporting biases are not
likely responsible for the longitudinal effects uncovered here. Extraversion, however,
is associated with increased sociability, which would be reflected in the interaction
quantity variable. This suggests, as discussed above, that personality may foster the
kind of socializing that must take place in early adulthood for long-term benefits to be
Regardless, these results must be interpreted with a degree of caution. We
cannot necessarily assume these results suggest personality caused a bias in
participant’s perceptions of college social activity, as personality was assessed 30
years later, in the current assessment. College socializing conceivably influenced
participants’ tendency to view the world through rose-colored glasses at age-50.
Although personality is said to remain stable over time, credible causal statements
would require an assessment of personality in college, along with an assessment of
the same midlife outcome measures of social and emotional adjustment. Only then
would it would be possible to further tease apart changes in those outcomes as a
function of both personality and social activity. Alas, the personality and outcome
measures were not collected (and in many cases, did not yet exist!) when these
participants were in college.
At midlife, virtually all measures (except neuroendocrine function) were self-
report. It would have been ideal to collect behavioral measures (e.g., a social
interaction task with a romantic partner or friend, or a medical evaluation by a doctor,
replete with blood work) and re-execute the experiencing sampling procedure using
the RIR. However, not only were participants spread across the world, it seemed
important to make participation as brief and simple as possible to achieve the lowest
possible rate of attrition.
A final limitation concerns the number of measures collected and thus the
number of analyses conducted. The chance of making Type I errors (mistakenly
rejecting the null hypothesis and assuming an effect to be real when in fact it is not)
was inflated because of the number of statistical analyses that were conducted.
Statistical corrections may reduce the likelihood of Type I errors, but were not done
here because these procedures tend to lower the power to detect real effects (possibly
resulting in Type II errors). A study of this type is somewhat exploratory by nature,
and correcting for the possibility of Type I errors risks the possibility of failing to find
real effects because the test is too stringent. Instead of jeopardizing the Type II error
rate by controlling for chance, observation of a consistent pattern of effects was used
to gauge the meaningfulness of these results. Inspection of Table 25 shows that six of
12 (50.0%) quantity effects reached significance and four of 12 (33.3%) quality
effects reached significance. In all, 41.7% of longitudinal effects were significant, in a
direction that supported the hypotheses. Moreover, the pattern of findings was
consistent for outcome type as well; for the most part, college social activity effects
emerged on social well-being (five of eight, or 62.5% of the effects were significant)
and emotional adjustment outcomes (four of six, or 66.7% of the effects were
marginally significant or better), and less so on physical health outcomes. The gold
standard, of course, would be to replicate these results with other samples.
Conclusions and Future Directions
This research is among the first to demonstrate that the features of one’s social
environment in early adulthood may have lasting implications for long-term social
and emotional well-being, and possibly health, although the results were substantially
weaker, and less reliable, for health outcomes. While many interesting and
provocative findings were uncovered, the study also raised further questions.
Longitudinally, although both structural social engagement (i.e., interaction
quantity) and functional support (i.e., interaction quality) in early adulthood predicted
later life outcomes, interaction quantity emerged as a relatively more robust and
consistent predictor of midlife outcomes over interaction quality. Other research has
identified both kinds of measures of support to be important to health and well-being,
and the findings of the current study raise questions of whether the relative impact of
structural and functional integration may vary a function of the time frame of
assessment. Perhaps during early adulthood, when socializing permeates the day-to-
day activities of the residential college student who is constantly surrounded by a sea
of potential interaction partners, frequent social activity is especially relevant and
important. However, as individuals move on from the residential hall to less socially
saturated living situations (e.g. with one or two housemates or a significant other), the
number of interaction partners likely shrinks and the quality of interactions with a few
selected close others may come to play a more important role.
The relative invariance of concurrent and longitudinal effects across sex raises
questions about the causes of sex differences in the strength and type of benefits
garnered from social life that have been identified in other research (reviewed above).
Sex effects that favor women typically manifest on well-being outcomes, which failed
to emerge for the current research – both men and women experienced social and
emotional adjustment benefits. Sex differences that favor men typically manifest on
health and mortality outcomes, however in the present research these results were
weak for both men and women. Two important questions that deserve further
exploration arise from these findings, one concerning the cause of sex differences in
other research, and the other having to do with the lack of health effects in this
research. A health-protective benefit associated with the relatively high subjective
socioeconomic status of this particular sample was offered as one explanation for the
lack of health findings at this stage. It will be particularly interesting to determine
whether health and well-being disparities vary as a function of early adult social life
as these individuals age and their health deteriorates. Moreover, it remains to be seen
whether sex differences will emerge in those effects, if indeed they do emerge.
This research offered a unique approach that contributed to our understanding
of the associations between social life and physical and mental health and wellness. It
examined richly detailed social activity data at a critical period in early adulthood,
and followed individuals into midlife. It illuminated the long-lasting impact of
socialization at an important life stage (college), and spoke to the importance of
considering the history from which a person’s social environment evolved. The
results suggest that people do, indeed, get by with a little help from their friends.
Abercrombie, H.C., Giese-Davis, J., Sephton, S., Epel, E.S., Turner-Cobb, J.M., &
Spiegel, D. (2004). Flattened Cortisol Rhythms in Metastatic Breast Cancer
Patients. Psychoneuroendocrinology, 29, 1082-1092.
Adler, N.E., Epel, E.S., Castellazzo, G., & Ickovics, J.R. (2000). Relationship of
subjective and objective social status with psychological and physiological
functioning: Preliminary data in healthy white women. Health Psychology, 19,
Almeida, D. M. (2005). Resilience and vulnerability to daily stressors assessed via
diary methods. Current Directions in Psychological Science, 14, 62-68.
Almeida, D. M., & Horn, M. C. (2004). Is daily life more stressful during middle
adulthood? In O. G. Brim & C. D. Ryff & R. C. Kessler (Eds.), How healthy
are we?: A national study of well-being at midlife (pp. 425-451). Chicago:
University of Chicago Press.
Argyle, M. (1999). Causes and correlates of happiness. In D. Kahneman, E. Diener,
& N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology (pp.
353-373). New York: Russell Sage.
Arnstein, R.L. (1984). Young adulthood: stages of maturity. In D. Offer & M.
Sabshin (Eds.). Normality and the life cycle (pp. 108-144). New York: Basic
Aron, A., Aron, E.N., and Smollan, D. (1992). Inclusion of other in self scale and the
structure of interpersonal closeness. Journal of Personality and Social
Psychology, 63, 596-612.
Baker, B., Paquette, M., Szalai, J.P., Driver, H., Perger, T., Helmers, K., O’Kelly, B.,
& Tobe, S. (2000). The influence of marital adjustment on 3-year left
ventricular mass and ambulatory blood pressure in mild hypertension.
Archives of Internal Medicine, 160, 3453-3458.
Baltes, P.B. (1997). On the incomplete architecture of human ontogeny: Selection,
optimization, and compensation as foundation of developmental theory.
American Psychologist, 52, 366-380.
Baltes, P.B., & Baltes, M.M. (1990). Psychological perspective on successful aging:
The model of selective optimization with compensation. In P.B. Baltes &
M.M. Baltes (Eds.), Successful aging: Perspectives from the behavioral
sciences (pp. 1-34). New York: Cambridge University Press.
Baumeister, R.F., & Leary, M.R. (1995). The need to belong: Desire for interpersonal
attachment as a fundamental human motivation. Psychological Bulletin, 117,
Berkman, L.F. (1995). The role of social relations in health promotion.
Psychosomatic Medicine, 57, 245-254.
Berkman, L.F., Glass, T., Brissette, I., & Seeman, T.E. (2000). From social
integration to health: Durkheim in the new millennium. Social Science and
Medicine, 51, 843-57.
Berkman, L.F., & Syme, S.L. (1979). Social networks, host resistance, and mortality:
a nine-year follow-up study of Alameda County residents. American Journal
of Epidemiology, 109, 186-204.
Birditt, K. S., Fingerman, K. L., & Almeida, D. M. (2005). Age differences in
exposure and reactions to interpersonal tensions: A daily diary study.
Psychology & Aging, 20, 330-340.
Blaney, P.H. (1986). Affect and memory: A review. Psychological Bulletin, 99, 229-
Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. New York: Basic Books.
Bowlby, J. (1979). The making and breaking of affectional bonds. London: Tavistock.
Bretherton, I, & Munholland, K.A. (1999). Internal working models in attachment
relationships: A construct revisited. In J. Cassidy & P.R. Shaver (Eds.),
Handbook of attachment: Theory, research, and clinical applications. New
York: The Guilford Press.
Buysse, D.J., Reynolds III, C.F., Monk, T.H., Berman, S.R., & Kupfer, D.J. (1989).
The Pittsburgh sleep quality index: a new instrument for psychiatric practice
and research. Journal of Psychiatric Research, 28, 193-213.
Byrne, D., and Clore, G.L. (1970). A reinforcement model of evaluative responses.
Personality: An International Journal, 1, 103-128.
Cacioppo, J.T., and Berntson, G.G. (1994). Relationship between attitudes and
evaluative space: A critical review, with emphasis on the separability of
positive and negative substrates. Psychological Bulletin, 115, 401-423.
Cacioppo, J.T., Ernst, J.M., Burleson, M.H., McClintock, M.K., Malarkey, W.B.,
Hawkley, L.C., Kowalewski, R.B., Paulsen, A., Hobson, J.A., Hudgahl, K.,
Spiegel, D., & Berntson, G.G. (2000). Lonely traits and concomitant
physiological processes: the MacArthur social neuroscience studies.
International Journal of Psychophysiology, 35, 143-154.
Cacioppo, J.T., Hawkley, L.C., Crawford, L.E., Ernst, J.M., Burleson, M.H.,
Kowalewski, R.B., Malarkey, W.B., Van Cauter, E., & Berntson, G.G. (2002).
Loneliness and health: Potential mechanisms. Psychosomatic Medicine, 64,
Caldwell, M.A., & Peplau, L.A. (1982). Sex differences in same-sex friendships. Sex
Roles, 8, 721-732.
Caprariello, P.A., & Reis, H.T. (2006). Development and implementation of a
perceived responsiveness scale. Poster presented at the biennial conference of
the International Association of Relationships Research, Rethymnon, Crete,
Carels, R.A., Sherwood, A., & Blumenthal, J.A. (1998). Psychosocial influences on
blood pressure during daily life. International Journal of Psychophysiology,
Carmichael, C.L., & Reis, H.T. (2005). Attachment, sleep quality, and depressed
affect. Health Psychology, 24, 526-531.
Carstensen, L.L. (1995). Evidence for a life-span theory of socioemotional selectivity.
Current directions in psychological science, 4, 151-156.
Carstensen, L.L., & Fredrickson, B.F. (1998). Socioemotional selectivity in healthy
older people and younger people living with the human immunodeficiency
virus: The centrality of emotion when the future is constrained. Health
Psychology, 17, 1-10.
Carstensen, L.L., Isaacowitz, D.M., & Charles, S.T. (1999). Taking time seriously: A
theory of socioemotional selectivity. American Psychologist, 54, 165-181.
Cassel, J.C. (1976). The contribution of the social environment to host resistance.
American Journal of Epidemiology, 104, 107-123.
Cobb, S. (1976). Social support as a moderator of life stress. Psychosomatic
Medicine, 38, 300-314.
Cohen, L.H., Towes, L.C., & Flocco, R. (1988). Effects of induced mood on self-
reported life events and perceived and received social support. Journal of
Personality and Social Psychology, 55, 669-674.
Cohen, S. (1991). Social supports and physical health: Symptoms, health behaviors
and infectious disease. In A. L. Greene, M. Cummings and K. H. Karraker
(Eds.), Life-span developmental psychology: Perspectives on stress and
coping. Hillsdale, NJ: Erlbaum.
Cohen, S., Doyle, W.J., Skoner, D.P., Rabin, B.S., & Gwaltney, J.M. (1997). Social
ties and susceptibility to the common cold. Journal of the American Medical
Association, 277, 1940-1944.
Cohen, S., & Hoberman, H.M. (1983). Positive events and social supports as buffers
of life change stress. Journal of Applied Social Psychology, 13, 99-125.
Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of perceived
stress. Journal of Health and Social Behavior, 24, 385-396.
Cohen, S., Mermelstein, R., Kamarck, T., & Hoberman, H. (1985). Measuring the
functional components of social support. In I. G. Sarason & B. R. Sarason
(Eds.), Social support: Theory, research and application. The Hague,
Holland: Martinus Nijhoff .
Cohen, S., & Wills, T.A. (1985). Stress, social support, and the buffering hypothesis.
Psychological Bulletin, 98, 310-357.
Cook, T.D., & Campbell, D.T. (1979). Quasi-experimentation: Design and analysis
for field settings. Boston: Houghton Mifflin.
Coyne, J.C., Rohrbaugh, M.J., Shoham, V., Sonnega, J.S., Nicklas, J.M., & Cranford,
J.A. (2001). Prognostic importance of marital quality for survival of
congestive heart failure. American Journal of Cardiology, 88, 526-529.
Danner, D.D., Snowden, D.A., & Friesen, W.V. (2001). Positive emotions in early
life longevity: Findings from the nun study. Journal of Personality and Social
Psychology, 80, 804-813.
Derogatis, L.S. (1977). The SCL-90 Manual I: Scoring, administration and
procedures for the SCL-90. Clinical Psychometric Research. Baltimore
Diamond, L.M., Hicks, A.M., & Otter-Henderson, K. (2008). Every time you go
away: Changes in Affect, Behavior, and Physiology Associated with Travel-
Related Separations from Romantic Partners. Journal of Personality and
Social Psychology, 95, 385-403.
Dickerson, S.S., & Kemeny, M.E. (2004). Acute stressors and cortisol responses: A
theoretical integration and synthesis of laboratory research. Psychological
Bulletin, 130, 355-391.
Diener, E., Emmons, R.A., Larsen, R.J., & Griffin, S. (1985). The satisfaction with
life scale. Journal of Personality Assessment, 49, 71-75.
Diener, E., Suh, E.M., Lucas, R.E., & Smith, H.L. (1999). Subjective well-being:
Three decades of progress. Psychological Bulletin, 125, 276-302.
Dienstbier, R.A. (1989). Arousal and physiological toughness: Implications for
mental and physical health. Psychological Review, 96, 84-100.
Eagly, A.H., & Wood. W. (1999). The origins of sex differences in human behavior:
Evolved dispositions versus social roles, American Psychologist, 54, 408-423.
Erikson, E.H. (1959). Identity and the life cycle. New York: W.W. Norton.
Field, D. (1999). Continuity and change in friendships in advanced old age: Findings
from the Berkeley older generation study. International Journal of Aging and
Human Development, 48, 81-83.
Fredrickson, B.L. (2000). Cultivating positive emotions to optimize health and well-
being. Prevention & Treatment, 3.
Fredrickson, B.L., & Levenson, R.W. (1998). Positive emotions speed recovery from
the cardiovascular sequelae of negative emotions. Cognition and Emotion, 12,
Funk, J.L., & Rogge, R.D. (2007). Testing the ruler with item response theory:
Increasing precision of measurement for relationship satisfaction with the
Couples Satisfaction Index. Journal of Family Psychology, 21, 572-583.
Gable, S.L., Reis, H.T., & Elliot A.J. (2003). Evidence for bivariate systems: An
empirical test of appetition and aversion across domains. Journal of Research
in Personality, 37, 349-372.
Glenn, N.D. (1975). The contribution of marriage to the psychological well-being of
males and females. Journal of Marriage and the Family, 37, 594-600.
Goldberg, L. R., Johnson, J. A., Eber, H. W., Hogan, R., Ashton, M. C., Cloninger, C.
R., & Gough, H. C. (2006). The International Personality Item Pool and the
future of public-domain personality measures. Journal of Research in
Personality, 40, 84-96.
Goldstein, H.H., Buss, A.H., Plomin, R., Rothbart, M.K., Thomas, A., Chess, S.,
Hinde, R.A., & McCall, R.B. (1987). Roundtable: What is temperament? Four
approaches. Child Development, 58, 505-529.
Goodwin, R.D., Hoven, C.W., Murison, R., & Hotopf, M. (2003). Association
between childhood physical abuse and gastrointestinal disorders and migraine
in adulthood. American Journal of Public Health, 93, 1065-1067.
Gump, B.B., Polk, D.E., Kamarck, T.W. & Shiffman, S.M. (2001). Partner
Interactions Are Associated With Reduced Blood Pressure in the Natural
Environment: Ambulatory Monitoring Evidence From a Healthy, Multiethnic
Adult Sample. Psychosomatic Medicine, 63, 423-433.
Harker, L., & Keltner, D. (2001). Expressions of positive emotion in women’s college
yearbook pictures and their relationship to personality and life outcomes
across adulthood. Journal of Personality and Social Psychology, 80, 112-124.
Hartup, W.W. (1993). Adolescents and their friends. In B. Laursen (Ed.), Close
friendships in adolescence (pp. 3-22). San Francisco: Jossey-Bass.
Hartup, W.W., & Stevens, N. (1997). Friendships and adaptation in the life course.
Psychological Bulletin, 121, 355-370.
Hawkley, L.C., & Cacioppo, J.T. (2007). Aging and loneliness: Downhill quickly?
Current Directions in Psychological Science, 16, 187-191.
Helgeson, V.S. (1991). The effects of masculinity and social support on recovery
from myocardial infarction. Psychosomatic Medicine, 53, 621-633.
Holt-Lunstad, J., Smith, T.B., & Layton, J.B. (2010). Social relationships and
mortality risk: A meta-analytic review. Public Library of Science Medicine,
7(7): e1000316. doi:10.1371/journal.pmed.1000316
House, J.S., Landis, K.R., & Umberson, D. (1988). Social relationships and health.
Science, 241, 540–545.
John, O.P., & Srivastava, S. (1999). The big-five trait taxonomy: History,
measurement, and theoretical perspectives. In L. Pervin and O.P. John (Eds.)
Handbook of personality: Theory and research. New York, NY: Guilford.
Kahneman, D., & Tversky, A. (1982). On the psychology of prediction. In D.
Kahneman, P. Slovic, & A. Tversky (Eds.), Judgment under uncertainty:
Heuristics and biases (pp. 48-67). New York: Cambridge University Press.
Keltener, D. (2003). Expression and the Course of Life: Studies of Emotion,
Personality, and Psychopathology from a Social-Functional Perspective. In P.
Ekman, J.J. Campos, and R.J. Davidson (Eds.) Emotions inside and out: 130
Years after Darwin's: The Expression of the Emotions in Man and Animals.
(pp. 222-243). New York, NY: New York University Press.
Kenny, D. A. (1975). Cross-lagged panel correlation: A test for spuriousness.
Psychological Bulletin, 82, 887-903.
Kessler, R.C., Kendler, K.S., Heath, A.C., Neale, M.C., Eaves, L.J. (1992). Social
support, depressed mood and adjustment to stress: A genetic epidemiologic
investigation. Journal of Personality and Social Psychology, 62, 257-272.
Kiecolt-Glaser, J. K., Glaser, R., Cacioppo, J. T., & Malarkey, W. B. (1998). Marital
stress: Immunologic, neuroendocrine, and autonomic correlates. Annals of the
New York Academy of Sciences, 840:649-655.
Kiecolt-Glaser, J.K., Malarkey, W.B., Chee, M., Newton, T., Cacioppo, J.T., Mao,
H., & Glaser, R. (1993). Negative behavior during marital conflict is
associated with immunological down-regulation. Psychosomatic Medicine,
Kiecolt-Glaser, J.K., & Newton, T.L. (2001). Marriage and health: His and hers.
Psychological Bulletin, 127, 472–503.
Kirschbaum, C., Kudielka, B.M., Gaab, J., Schommer, N.C., & Hellhammer, D.H.
(1999). Impact of gender, menstrual cycle phase, and oral contraceptives on
the activity of the hypothalamus-pituitary-adrenal axis. Psychosomatic
Medicine, 61, 154-162.
King, K.B., Reis, H.T., Porter, L.A., & Norsen, L.H. (1993). Social support and long-
term recovery from coronary artery surgery: Effects on patients and spouses.
Health Psychology, 12, 56-63.
Larsen, J.T., McGraw, A.P., & Cacioppo, J.T. (2001). Can people feel happy and sad
at the same time? Journal of Personality and Social Psychology, 81, 684-696.
Levenson, R.W., Carstensen, L.L, & Gottman, J.M. (1993). Long-term marriage:
Age, gender, and satisfaction. Psychology and Aging, 2, 301-313.
Levinson, D.J. (1978). The seasons of a man’s life. New York: Knopf.
Lydon, J.E., Jamieson, D.W., & Holmes, J.G. (1997). The meaning of social
interactions in the transition from acquaintanceship to friendship. Journal of
Personality and Social Psychology, 73, 536-548.
Lyubomirsky, S., King, L., & Diener, E. (2005). The benefits of frequent positive
affect: Does happiness lead to success? Psychological Bulletin, 131, 803-855.
Maier, E. H., & Lachman, M. E. (2000). Consequences of early parental loss and
separation for health and well-being in midlife. International Journal of
Behavioral Development, 24, 183-189.
Malarkey, W., Kiecolt-Glaser, J.K., Pearl, D., & Glaser, R. (1994). Hostile behavior
during marital conflict alters pituitary and adrenal hormones. Psychosomatic
Medicine, 56, 41-51.
Manuck, S.B. (1994). Cardiovascular reactivity in cardiovascular disease: “Once
more unto the breach.” International Journal of Behavioral Medicine, 1, 4-31.
McEwen, B.S. (1998). Protective and Damaging Effects of Stress Mediators. New
England Journal of Medicine, 338, 171-179.
McEwen, B S., & Stellar, E. (1993). Stress and the individual. Mechanisms leading to
disease. Archives of Internal Medicine, 153, 2093-101.
Mendes, W.B., Reis, H.T., Seery, M.D., & Blascovich, J. (2003). Cardiovascular
correlates of emotional expression and suppression: Do content and gender
context matter? Journal of Personality and Social Psychology, 84, 771-792.
Milardo, R.M., Johnson, M.P., & Huston, T.L. (1983). Developing close
relationships: Changing patterns of interaction between pair members and
social networks. Journal of Personality and Social Psychology, 44, 964-976.
Murray, S.L, Rose, P., Bellavia, G.M., Holmes, J.G., & Kusche, A.G. (2002). When
rejection stings: How self-esteem constrains relationship-enhancement
processes. Journal of Personality and Social Psychology, 83, 556-573.
National Center for Health Statistics. National Health and Nutrition Examination
Survey, 2002. Public-use data file and documentation.
National Center for Health Statistics. National Health and Nutrition Examination
Survey, 2005. Public-use data file and documentation.
Neugarten, B.L., Moore, J.W., & Lowe, J.C. (1965). Age norms, age constraints, and
adult socialization. American Journal of Sociology, 70, 710-717.
Nikiforov, S.V., & Mamaev, V.B. (1998). The development of sex differences in
cardiovascular disease mortality: a historical perspective. American Journal of
Public Health, 88, 1348-1353.
Orlofsky, J.L. (1976). Intimacy status: Relationship to interpersonal perception.
Journal of Youth and Adolescence, 5, 73-88.
Orlofsky, J.L., & Ginsberg, S.D. (1981). Intimacy status: Relationship to affect
cognition. Adolescence, 16, 91-100.
Orth-Gomer, K., Rosengren, A., & Wilhelmsen, L. (1993). Lack of social support and
incidence of coronary heart disease in middle-aged Swedish men.
Psychosomatic Medicine, 55, 37-43.
Peplau, L.A., & Perlman, D. (2000). Loneliness. A sourcebook of current theory,
research, and therapy. New York, NY: Wiley Interscience.
Pietromonaco, P.R., & Feldman-Barrett, L. (1997). Working models of attachment
and daily social interactions. Journal of Personality and Social Psychology,
Pressman, S.D., & Cohen, S. (2005). Does positive affect influence health?
Psychological Bulletin, 131, 925-971.
Pruessner, J.C., Kirschbaum, C., Meinlschmid, G., and Hellhammer, D.H. (2003).
Two formulas for computation of the area under the curve represent measure
of total hormone concentration versus time-dependent change.
Psychoneuroendocrinology, 28, 916-931.
Radloff, L.S. (1977). The CES-D scale: A self-report depression scale for research in
the general population. Applied Psychological Measurement, 1, 385-401.
Reis, H.T., Collins, W.A., & Berscheid, E. (2000). The relationship context of human
behavior and development. Psychological Bulletin, 126, 844-872.
Reis, H.T., & Gable, S.L. (2000). Event-sampling and other methods for studying
everyday experience. In H.T. Reis & C.M. Judd (Eds.), Research methods in
social and personality psychology (pp. 190-222). Cambridge, UK: Cambridge
Reis, H.T., Lin, Y., Bennett, E.S., & Nezlek, J. (1993). Change and consistency in
social participation during early adulthood. Developmental Psychology, 29,
Reis, H.T., Nezlek, J., & Wheeler, L. (1980). Physical attractiveness in social
interaction. Journal of Personality and Social Psychology, 38, 604-617.
Reis, H.T., & Wheeler, L. (1991). Studying social interaction with the Rochester
interaction record. Advances in experimental social psychology, 24, 269-318.
Reis, H.T., Wheeler, L., Spiegel, N., Kernis, M., Nezlek, J., & Perri, M. (1982).
Physical attractiveness in social interaction: II. Why does appearance affect
social experience? Journal of Personality and Social Psychology, 43, 979-
Roberto, K.A. (1997). Qualities of older women’s friendships: Stable or volatile?
International Journal of Aging and Human Development, 44, 1-14.
Robles, T.F., Shaffer, V.A., Malarkey, W.B., & Kiecolt-Glaser, J.K. (2006). Positive
behaviors during marital conflict: Influences on stress hormones. Journal of
Social and Personal Relationships, 23, 305-325.
Rook, K.S. (1984). The negative side of social interaction: Impact on psychological
well-being. Journal of Personality and Social Psychology, 46, 1097-1108.
Ryff, C.D., & Keyes, C.L. (1995). The structure of psychological well-being
revisited. Journal of Personality and Social Psychology, 69, 719-727.
Ryff, C.D., Love, G.D., Urry, H.L., Muller, D., Rosenkranz, M.A., Friedman, E.M.,
Davidson, R.J., & Singer, B. (2006). Psychological well-being and ill-being:
Do they have distinct or mirrored biological correlates? Psychotherapy and
Psychosomatics, 75, 85-95.
Russell, J. (1996). UCLA loneliness scale (version 3): Reliability, validity, and factor
structure. Journal of Personality Assessment, 66, 20-40.
Sapolsky, R.M. (1990). Stress in the wild. Scientific American, 262, 116-123.
Sapolsky, R.M. (1992). Stress and neuroendocrine changes during aging.
Generations: Journal of the American Society on Aging, 16, 35-38.
Sapolsky, R.M. (1996). Why stress is bad for your brain. Science, 273, 749-750.
Schaefer, M.T., & Olson, D.H. (1981). Assessing intimacy: The PAIR Inventory.
Journal of Marital and Family Therapy, 7, 47-60.
Schmidt-Reinwald, A., Pruessner, J.C., Hellhammer, D.H., Federenko, I., Rohleder,
N., Schumeyer, T.H., & Kirschbaum, C. (1999). The cortisol response to
awakening in relation to different challenge tests and a 12-hour cortisol
rhythm. Life Sciences, 64, 1653-1660.
Schwarz, N., Groves, R.M., & Schuman, H. (1998). Survey methods. In D.T. Gilbert,
S.T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (Vol. 1,
4th ed., p. 143-179). New York: McGraw Hill.
Seeman, T.E. (1996). Social ties and health: The benefits of social integration. Annals
of Epidemiology, 6, 442-451.
Seeman, T.E., Singer, B.H., Ryff, C.D., Love, G.D., & Levy-Storms, L. (2002).
Social relationships, gender, and allostatic load across two age cohorts.
Psychosomatic Medicine, 64, 395-405.
Seeman, T.E., McEwen, B.S., Rowe, J.W., & Singer, B.H. (2001). Allostatic load as a
marker of cumulative biological risk: MacArthur studies of successful aging.
Proceedings of the National Academy of Sciences, 98, 4770-4775.
Shaver, P.R., & Mikulincer, M. (2002). Attachment-related psychodynamics.
Attachment and Human Development, 4, 133-161.
Skowronski, J.J., Betz, A.L., Thompson, C.P., & Shannon, L. (1991). Social memory
in everyday life: Recall of self-events and other-events. Journal of Personality
and Social Psychology, 60, 831-843.
Sterling, P., & Eyer, J. (1981). Biological basis of stress-related mortality. Social
Science & Medicine, 15E, 3-42.
Stewart, J. & Seeman, T. (2000). Salivary cortisol measurement. Retrieved August
23, 2006 from the John D. and Catherine T. MacArthur Research Network on
Socioeconomic Status and Health Web site:
Sullivan, H.S. (1953). The interpersonal theory of psychiatry. New York, NY: W. W.
Norton & Co.
Taris, T. W. (2000). A primer in longitudinal data analysis. London: Sage.
Taylor, S.E., Klein, L.C., Lewis, B.P, Gruenwald, T.L., Gurung, R.A.R., &
Updegraff, J.A. (2000). Biobehavioral responses to stress in females: Tend-
and-befriend, not fight-or-flight. Psychological Review, 107, 411-429.
Tugade, M.M., & Fredrickson, B.L. (2004). Resilient individual use positive
emotions to bounce back from negative emotional experiences. Journal of
Personality and Social Psychology, 86, 320-333.
Turner-Cobb, J.M., Sephton, S.E., Koopman, C., Blake-Mortimer, J., & Spiegel, D.
(2000). Social support and salivary cortisol in women with metastatic breast
cancer. Psychosomatic Medicine, 62, 337-345.
Uchino, B.N., Cacioppo, J.T., & Kiecolt-Glaser, J.K. (1996). The relationship
between social support and physiological processes: a review with emphasis
on underlying mechanisms and implications for health. Psychological
Bulletin, 119, 483-531.
Uchino, B.N., Kiecolt-Glaser, J.K., & Cacioppo, J.T. (1992). Age-related changes in
cardiovascular responses as a function of a chronic stressor and social support.
Journal of Personality and Social Psychology, 63, 839-846.
Uchino, B.N., Holt-Lunstad, J., Uno, D., & Flinders, J.B. (2001). Heterogeneity in the
social networks of young and older adults: Prediction of mental health and
cardiovascular reactivity during acute stress. Journal of Behavioral Medicine,
Umberson, D. (2002). Gender, marital status and the social control of health behavior.
Social Science & Medicine, 34, 907-917.
Umberson, D., Chen, M.D., House, J.S., Hopkins, K., & Slaten, E. (1996). The effect
of social relationships on psychological well-being: Are men and women
really so different? American Sociological Review, 61, 837-857.
United Nations, Department of Economic and Social Affairs, Population Division
(2009). World Marriage Data 2008.
Ware, J.E., & Sherbourne, C.D. (1992). The MOS 36-item short-form health survey
(SF-36). I. Conceptual framework and item selection. Medical Care, 6, 473-
Watson, D., & Clark, L.A. (1994). The PANAS-X: Manual for the Positive and
Negative Affect Schedule-Expanded Form. Unpublished manuscript,
University of Iowa, Iowa City, IA.
Wenger, G.C., & Burholt. V. (2004). Changes in level of social isolation and
loneliness among older people in a rural area: A twenty-year longitudinal
study. Canadian Journal on Aging, 23, 115-127.
Wheeler, L., & Nezlek, J. (1977). Sex differences in social participation. Journal of
Personality and Social Psychology, 35, 742-754.
Wheeler, L., Reis, H.T., Nezlek, J. (1983). Loneliness, social interaction, and sex
roles. Journal of Personality and Social Psychology, 45, 943-953.
White, R.W. (1966). Lives in progress. New York: Holt, Rienhart & Winston.
Williams, D.R., & Collins, C. (1995). US socioeconomic and racial differences in
health: Patterns and explanations. Annual Review of Sociology, 21, 349-386.
Wood, W., Rhodes, N., & Whelan, M. (1989). Sex differences in positive well-being:
A consideration of emotional style and marital status. Psychological Bulletin,
Wust, S., Federenko, I., Hellhammer, D.H., & Kirschbaum, C. (2000). Genetic
factors, perceived chronic stress, and the free cortisol response to awakening.
Psychoneuroendocrinology, 25, 707-720.
1. The total sample size from which we are recruiting (n = 222) differs from the sum
of the number of participants in the three individual studies (n = 248) because 26
individuals (19F, 7M) participated in both studies II and III. Half of these 26
participants were randomly assigned to sample II (10F, 3M), and the other half to
sample III (9F, 4M) in order to determine which college data (sample II or III)
would be used in longitudinal analyses.
2. One transgendered participant (male in college) had to be dropped from all
analyses of sex because a comparison with a group consisting of one member is
3. Oblique solutions were chosen to best capture the factor structure of the variables
while not forcing principal components to be mutually independent. Given the
types of variables included in each principle component analysis, modest
correlations among factors were expected.
4. Although this cortisol variable is termed area under the curve with respect to
increase, because cortisol values tend to decrease over the day, many of the
values for this calculation become negative, and represent the decrease in cortisol
over the course of the day (see Pruessner et al., 2003 for a further explanation).
5. A set of analyses was undertaken to test for effects of the interaction between
college interaction quantity and quality, as well as the three-way interaction with
sex. Neither the Quantity X Quality interaction, nor the Sex X Quantity X Quality
had significant effects on midlife social well-being, emotional adjustment, or
physical health outcomes. These variables were thus deleted from the model and
are not discussed further.
6. The transgendered participant was again omitted from analyses including sex
(Model II). Inclusion of this participant did not meaningfully alter the quantity
and quality main effects reported in Model I throughout.
7. College data for the additional 42 of the original 222 participants were excluded
because those participants were never located, and mortality status could not be
Table 1. Summary of function and measurement of three physiological systems
Major Functions & Organs
Transport oxygen to, and remove
carbon dioxide from cells and
organs in the body to maintain
cell and organ function.
Heart: generates force for
Arteries, veins, & capillaries (the
vasculature): conduit for the
pumping of the heart
Heart rate: beats per minute
of heart (increased by
sympathetic nervous system,
decreased by parasympathetic
Systolic blood pressure (SBP):
force of blood against arterial
wall during ventricular
contraction (highest pressure)
Diastolic blood pressure
(DBP): force of blood against
arterial wall during ventricular
relaxation (lowest pressure)
Defend body against disease
Thymus, bone marrow, lymph
nodes, spleen, tonsils, appendix
Immune cell counts: quantity
of immune cells (helper T,
suppressor T, NK cells)
Balance of immune cells: ratio
of types of immune cells such
as helper T cells to suppressor
Immune cell performance:
proliferation of immune cells
in response to antigens
Regulate cardiovascular system,
metabolize stored glucose and fat,
down regulate immune system
Adrenal cortex, adrenal medulla,
(Note: the hypothalamus,
pituitary, and adrenal cortex
make up the HPA axis)
& norepinephrine, produced
by adrenal medulla), CRH
(released by hypothalamus),
ACTH (released by pituitary),
cortisol (released by adrenal
Catecholamines and cortisol
follow daily cycle peaking at
awakening and gradually
declining over the day
Table 2. Summary of previous samples and recruitment for current study
Year of initial
Age at initial
n recruited for
Note: Sample size reflects the number of individuals who participated in each original study. Parenthetically reported is the
number of unique individuals assigned to that study, with half of the 26 individuals who participated in both the second and
third studies being randomly assigned to each sample.
Table 3. Comparison of recruited versus unrecurited participants on college social
College interaction quantity
Predictor Unrecruited M Recruited M t df p
Overall Quantity -.01 .06 -1.34 207 ns
TPD 327.44 353.10 -1.50 207 ns
#PD 5.07 5.45 < 1 207 ns
Same Sex .00 .00 < 1 207 ns
TPD 158.92 160.09 < 1 207 ns
#PD 3.61 3.57 < 1 207 ns
Opposite Sex -.12 .04 -1.38 207 ns
TPD 75.71 93.41 -1.66 207 = .10
#PD 1.55 1.75 -1.04 207 ns
College interaction quality
Predictor Unrecruited M Recruited M t df p
Overall Quality -.06 .04 < 1 207 ns
INT 3.52 3.55 < 1 207 ns
SAT 3.62 3.82 -1.13 207 ns
Same Sex -.06 .04 < 1 207 ns
INT 3.49 3.54 < 1 207 ns
SAT 3.68 3.84 -1.01 207 ns
Opposite Sex -.03 .02 < 1 207 ns
INT 4.02 3.99 < 1 207 ns
SAT 3.58 3.74 < 1 207 ns
Note: N = 222. Missing college data for 13 Ss. Recruited n with college data = 129,
unrecruited n with college data = 80.
Table 4. Factor loadings for individual social relationship quality scales on three
principle components with oblimin rotation.
Relationship Friendship Social
Item Quality Quality Network Communalities
KPSS-R .898 .236 .173 .807
IOS-R .878 .182 .220 .775
PAIR-R .460 .892 .283 .855
CSI .957 .241 .231 .917
PPR .936 .271 .144 .881
KPSS-F .087 .818 .094 .697
IOS-F .237 .760 .334 .598
PAIR-F .210 .941 .289 .888
NTWK-DV .200 .241 .930 .866
NTWK-SZ .183 .266 .922 .850
Eigenvalues 4.397 2.286 1.451
Note: Bolded values indicate the item belongs on that principle component. KPSS-R:
perceived support from romantic partner, IOS-R: inclusion of other in self with
romantic partner, PAIR-R: intimacy with romantic partner, CSI: satisfaction with
romantic relationship, PPR: perceived partner responsiveness, KPSS-F: perceived
support from core friend group, IOS-F: inclusion of other in self with core friend
group, PAIR-F: intimacy with core friend group, NTWK-DV: social network
diversity, NTWK-SZ: social network size
Table 5. Factor loadings for individual emotional adjustment scales on three principle
components with oblimin rotation.
Positive Negative Self-
Item affectivity affectivity actualization Communalities
SWLS .737 -.526 .420 .586
LONELY -.867 .437 -.307 .764
SELF-ACC .806 -.605 .680 .809
POS-REL .876 -.353 .430 .774
PA .647 -.444 .544 .508
NA -.394 .880 -.276 .777
CESD -.612 .840 -.446 .770
SCL90D -.628 .816 -.395 .741
SCL90A -.215 .789 -.177 .658
PSS -.531 .847 -.448 .752
PRS-GRTH .463 -.281 .803 .658
PURP .601 -.387 .781 .685
AUT .221 -.267 .844 .749
ENV-MAST .653 -.661 .585 .650
Eigenvalues 7.283 1.574 1.024
Note: Bolded values indicate the item belongs on that principle component. SWLS:
satisfaction with life scale; LONELY: UCLA-Loneliness, SELF-ACC: Ryff self-
acceptance, POS-REL: Ryff positive relations with others, PA: positive affect, NA:
negative affect, CESD: center for epidemiological studies – depression, SCL90D:
SCL90 depression subscale, SCL90A: SCL90 anxiety subscale, PSS: perceived stress
scale, PERS-GRTH: Ryff personal growth, PURP: Ryff purpose in life, AUT: Ryff
autonomy, ENV-MAST: Ryff environmental mastery
Table 6: Factor loadings for individual physical health scales on three principle
components with oblimin rotation.
Daily Cardiac Cardiac
Item Symptomatology Health I Health II Communalities
CHIPS .823 -.046 -.090 .688
SF-36 PHYS -.713 -.186 -.085 .550
RESPIR .773 .035 -.013 .598
CARDIAC .099 .467 .419 .404
SYSBP .114 .869 .088 .776
DIABP -.019 .890 -.054 .796
CHOLEST -.225 -.030 .687 .524
BMI .301 .336 .649 .624
ENDOCRN .161 -.021 .663 .466
Eigenvalues 2.567 1.607 1.252
Note: Bolded values indicate the item belongs on that principle component. CHIPS:
daily symptoms, SF-36 PHYS: physical health subscales of the SF-36, RESPIR: prior
respiratory diagnoses, CARDIAC: prior cardiac diagnoses, SYSBP: systolic blood
pressure, DIABP: diastolic blood pressure, CHOLEST: cholesterol level, BMI: body
mass index, ENDOCRN: prior endocrine system diagnoses
Table 7. Effects of original sample and year in college (freshman vs. senior) on midlife social well-being
X Sex F
Note: Degrees of freedom vary from 1,105 to 2,125
**p ! .01, *p ! .05, †p ! .10
Table 8. Effects of original sample and year in college (freshman vs. senior) on midlife emotional adjustment
X Sex F
Note: Degrees of freedom vary from 1,119 to 2,122
**p ! .01, *p ! .05, †p ! .10
Table 9. Effects of original sample and year in college (freshman vs. senior) on midlife physical health
X Sex F
Cardiac Health I
Cardiac Health II
Note: DF vary from 1,68 to 2,126
**p ! .01, *p ! .05, †p ! .10
Table 10. Effects of original sample and year in college (freshman vs. senior) on neuroendocrine function
X Sex F
Note: Degrees of freedom vary from 1,76 to 2,76
**p ! .01, *p ! .05, †p ! .10
Table 11. Concurrent correlations among midlife social, emotional, and physical health composite variables.
Social Well Being
Social Well Being
Cardiac Health I
Cardiac Health II
Note: **p ! .01, *p ! .05, †p ! .10
Table 12. Concurrent correlations among midlife social well being, emotional adjustment, physical health and neuroendocrine
composites and individual measures of social well being and emotional adjustment.
Social Well Being
Social Well Being
Note: **p ! .01, *p ! .05, †p ! .10
Table 13. Concurrent correlations among midlife social well being, emotional adjustment, physical health and neuroendocrine
composites and individual measures of physical health and neuroendocrine function.
Social Well Being
Note: **p ! .01, *p ! .05, †p ! .10
Table 14. Concurrent correlations among midlife social well being, emotional adjustment, physical health, and neuroendocrine
function composites for men (below the diagonal) and women (above the diagonal)
Social Well Being
Social Well Being
Cardiac Health I
Cardiac Health II
**p ! .01, *p ! .05, †p ! .10