AMERICAN

Journal of Epidemiology
Formerly AMERICAN JOURNAL OF HYGIENE
© 1983 by Tlie Johns Hopkins University School of Hygiene and Public Health

VOL. 117 MAY 1983 NO. 5

Reviews and Commentary

THE EPIDEMIOLOGIC EVIDENCE FOR A RELATIONSHIP BETWEEN
SOCIAL SUPPORT AND HEALTH
W. EUGENE BROADHEAD,1-1 BERTON H. KAPLAN,1 SHERMAN A. JAMES, 1
EDWARD H. WAGNER,1 VICTOR J. SCHOENBACH,1 ROGER GRIMSON,1 SIEGFRIED HEYDEN,'
GOSTA TIBBLIN' AND STEPHEN H. GEHLBACH1

The role of the social environment in reported symptoms and illness behavior
host resistance had been studied for some through a variety of chronic and infec-
time when in the mid 1970s, three major tious diseases, pregnancy outcome, psy-
papers were published which reviewed chiatric morbidity, childhood develop-
the mounting evidence that "social sup- ment, suicide, accidents, recovery from
port" had both a direct positive effect on illness, and death from a number of
health status and served as a buffer or chronic diseases. The measures of social
modifier of the effects of psychosocial and support were as varied as the number of
physical stress on the mental and physi- investigators and included the presence of
cal health of the individual (1-3). The litter mates in animal studies, social dis-
study designs reviewed included the organization, rapid social change, accul-
whole gamut of scientific inquiry— turation, morale, presence of a confidant,
animal experiments and ecologic, cross- presence of a family member, children
sectional, case-control, cohort and ran- having been wanted, family competence,
domized controlled trial studies. The and emotionally supportive intervention
ameliorative and protective effects of so- by clinicians. Likewise, in studies of the
cial support were reported for numerous buffering effect of social support, the mea-
disease outcomes ranging from self- sures of stress ranged from indirect proxy
1
measures and presumed stressful situa-
Duke-Watts Family Medicine Program, Dur- tions to quantitative measures (e.g., the
ham, NC.
1
University of North Carolina School of Public Schedule of Recent Experience (4)).
Health, Chapel Hill, NC.
1
The attention to this area created by
Department of Community and Family Medi- these reviews resulted in an outpouring of
cine, Duke University Medical Center, Durham, NC.
4
Department of Social Medicine, Akademiska publications of social support research
sjukhoset, Uppsala, Sweden. since the mid 1970s. The quantity and di-
Reprint requests to Dr. Berton H. Kaplan, De- versity of this new research necessitates
partment of Epidemiology, School of Public Health,
University of North Carolina, Chapel Hill, NC and permits a careful review and re-
27514. evaluation of the relationships between
521
522 BROADHEAD ET AL.

social support and health suggested ear- ity—especially with respect to the asso-
lier. ciation between low social support and
Cross-sectional studies abound and mental illness or psychologic symptoms.
show a direct association between a vari- Prospective cohort studies have con-
ety of social support measures and de- firmed the direct beneficial effects of vari-
pression (5 — 7), anxiety (5, 6), other ous forms of social support on global men-
psychologic or psychiatric symptoms tal health (23), incidence of depressive
(8-13), physical or somatic symptoms (5, symptoms (24, 25), recovery from a unipo-
7, 8, 14), self-reported use of health ser- lar depressive episode (26), psychologic
vices (14) and blood pressure (15). There distress (27, 28), psychologic "strain" (29),
are cross-sectional studies, however, which physical symptoms (24) and all-causes
fail to find these direct effects for both mortality (30-32). Two of these studies
psychologic and physical variables (16). show that the effect of social support is
And there is some variation in this as- greatest in the presence of social stressors
sociation by sex; two studies documented (effect modification): one for serum cho-
a negative association between social lesterol levels (24) and one for psychologic
support and number of symptoms for symptoms (28). One study (23), however,
women, but not for men (5, 8). The statis- did not detect effect modification despite
tical interaction of social support and the presence of direct effects. Another (25)
stressful life events (or other stressors)— detected no effect on physical symptoms,
and thus a hypothesized effect modifi- despite the positive association with
cation—has been confirmed by a number psychologic status.
of these cross-sectional studies that had Other authors (33, 34) have reviewed
also found a direct effect of social support numerous clinical and community-based
independent of interaction terms (5, 6, 8, intervention programs which have unfor-
9, 14), and by one study looking only at tunately not been properly evaluated
the buffering hypothesis (17). One study with pre- and post-intervention measures
supported the buffering hypothesis but of perceived social support and mental or
did not find significant direct effects (16). physical health outcomes. Three ade-
And one study found no interactive effect, quately designed clinical intervention
although a significant direct effect was programs do bear mentioning. The first,
present (12). by Pless and Satterwhite (35) involved a
Several retrospective and case-control randomized controlled clinical trial of lay
studies have also shown a direct effect of family counselors and families of children
social supports in various health/disease with chronic diseases. After one year of
states: psychologic adjustment to divorce follow-up, the percentage of children with
(18), physical and emotional recovery improved psychologic status was greater
after an automobile accident (19), and in the treatment group vs. the control (60
nonpsychotic psychiatric disorders (20, vs. 41 per cent, respectively). Gottlieb (36)
21). One cohort study which evaluated the used a randomized control design to eval-
presence of social supports retrospectively uate the effect of physician-lead support
detected no direct effect, but a significant groups for parents experiencing the stress
interactive effect (22). But, despite their of having their first child. He was able
contribution to our understanding of the to document an increase in the amount of
relationship of the social environment to support received in the patients' own sup-
health and disease, these, like the cross- port networks, outside the groups, but
sectional studies, are flawed by their in- could not demonstrate a reduction in
ability to address the direction of causal- subjective ratings of stress or an im-
 SOCIAL SUPPORT AND HEALTH 523
provement in sense of well-being. The baseline status into account. This can be
third study, by Sosa et al. (37) used sup- accomplished by stratified analysis or by
portive lay companions for women during multivariable methods which include
labor, also in a randomized controlled de- baseline status as an independent vari-
sign. Controls had higher complication able. It is only after such analyses that the
rates (cesarean section, meconium stain- causal direction between low social sup-
ing, etc.), but even when comparing only port and health outcome can be inferred;
uncomplicated deliveries, the experimen- unfortunately, the bulk of the studies we
tal group had markedly shorter labors have reviewed to this point have not met
than controls (8.8 vs. 19.3 hours), were this criterion.
more often awake after delivery and There are five recent exceptions. Wil-
stroked, smiled at and talked to their liams et al. (23) demonstrated a signifi-
babies more than did control mothers. cant relationship between baseline social
We have reviewed a large number of support and mental health at one year of
reported associations, but we should not follow-up in a multiple regression model
hastily conclude that a causal relation- controlling for baseline mental health as
ship exists between low social support and an independent variable (standardized
unfavorable health outcomes based on the regression coefficient of 0.12). The mag-
sheer weight of numbers. Rather, we nitude of the association was somewhat
should take the advice of Austin Bradford decreased from the model lacking base-
Hill (38) and closely examine the charac- line status as an independent variable
teristics of this association before making (standardized coefficient of 0.19), but it
interpretations of causality. Hill proposed was still significant (p < 0.01). Holahan
eight criteria to be considered when infer- and Moos (27) showed similar minor re-
ring causality: temporality, strength, ductions in measure of effect of work-
consistency, biologic gradient, biologic related support on men and family-
plausibility, coherence, experimental/ related support on women. However, a
intervention and specificity of outcome. striking effect occurred for work-related
We discuss current social support re- support and women; in this case, control-
search with respect to these and other ling for baseline psychologic distress de-
criteria and summarize this discussion in creased the presumed (uncontrolled) mea-
table 1. sure of effect from a significant value (r =
Temporality. In traditional prospective -0.225, p = 0.02) to an insignificant level
cohort studies of specific chronic diseases, (r = 0.005, nonsignificant). Turner (28),
diseased persons are excluded from the using similar statistical methods, found a
original sample and the nondiseased significant effect of social support on the
cohort is followed and evaluated for dis- psychologic well-being of new mothers
ease incidence over time. However, in after controlling for baseline status.
studies evaluating the incidence of changes Berkman and Syme (30) used stratifica-
in global measures such as mental health tion analysis to control carefully for age,
or general physical health status, dichoto- baseline health status, health practices,
mization of subjects into diseased vs. non- and utilization of preventive health ser-
diseased groups would be arbitrary. But vices in their study of low social support
since the mental health or physical status and mortality and found the relationship
of a person at one point in time can be pre- independent of these baseline risk factors.
sumed to affect his or her mental health And, most recently, Blazer (32) demon-
and physical status (including mortality) strated a strong association between per-
into the future, it is necessary to take ceived social support and 30-month mor-
524 BROADHEAD ET AL.

TABLE 1
Characteristics of the association between social support and health

Temporality:
Poor social support precedes adverse psychologic outcomes and mortality.
Strength:
Social support explains from 1.0-7.0% of the variance in psychologic outcomes. The relative risk of
mortality given poor social support is in the range of 1.5-3.5.
Consistency:
There is a similar direction and magnitude of effect across all major study designs and across a wide
variety of age, sex, race, ethnic, and health status groups. But the effect is greater for women than for
men in most studies.
Biologic gradient:
There is an apparent increase in numbers of physical and psychologic symptoms and mortality with
incremental increases in numbers or frequency of social contacts. The relationship is less clear for
perceived qualitative measures of social support.
Biologic plausibility:
Experimental evidence (animal and human) suggests neuroendocrine mechanisms, possibly mediated by
/3-endorphin, which might explain both the proposed direct and stress-modifying effects of social support.
Coherence:
Social support theory is supported by studies in ethology, and existing psychosocial theory and biologic
evidence can be used to explain the effect of social support at six or more points in the proposed causal
chain between exposure and disease.
Experimental^ intervention:
Social support intervention has improved psychologic outcome of chronically ill children and pregnancy
outcome of women in labor. Otherwise there is a dearth of adequately evaluated intervention.
Specificity of outcome:
The wide number of physical and psychologic outcomes associated with variations in social support are
consistent with cognitive and neuroendocrine mechanisms of effect.
Measurement of exposure:
A wide range of definitions of social support have been used. Factor analytic studies suggest constructual
differences between measures of quality vs. quantity. Inappropriate summary indexes and contamina-
tion of indexes by nonsupport variables plague the literature.
Determinants of social support:
A large number of environmental and individual characteristics interact to produce a person's social
support system at any one point in time. Health outcomes affect each of these determinants. (Social
support is not merely an environmental exposure.)
Dynamics of social support:
The nature of all these determinants changes with sequential role changes and other life events as an
individual proceeds through the life cycle. (Social support is not independent of life events).

tality in an elderly group of community perceived health and symptom status

residents, controlling for baseline physi- (14). Unfortunately, we can predict such
cal health, activities of daily living, de- associations in advance if we presume
pressive symptoms, age and six other po- that mental or physical illness has a di-
tential confounders. rect effect on access to social supports or
Strength. The strength of the signifi- perception of socially supportive behavior
cant associations of low social support by others. Moving on to prospective cohort
with poor health status reported in the studies we find multiple partial correla-
numerous cross-sectional and retrospec- tion coefficients ranging from 0.17 to 0.33
tive studies cited above ranges from r = for psychologic variables, and relative
0.14 to r = 0.55 for psychiatric symptoms, risks of 2.6 for mortality (31) and 2.72 for
and, for those using tabular data, preva- complications of pregnancy in a high-
lence ratios range from 1.44 for depres- stress group of mothers (40). But, as was
sive symptoms (39) to 2.54 for self- demonstrated in the previous section, the
 SOCIAL SUPPORT AND HEALTH 525
apparent strength of these associations have discussed above, in whether the ef-
will be reduced—even in prospective fect of social support is direct or indirect
studies controlling for other risk factors— (effect modification) or both (see "Biologic
when baseline health status is added to plausibility" below), and in the effect by
the model as an independent variable. In sex. However, where sex differences have
those studies which meet this latter cri- been examined, the results are consistent;
terion the multiple partial correlation co- the effect of social support on health out-
efficients range in magnitude from 0.11 to come is always greater for females (5, 8,
0.27, explaining from 1-7 per cent of the 27, 30).
variance in mental health outcome. Simi- Biologic gradient. A simple cause-effect
larly, Berkman and Syme (30) demon- relationship is more tenable if we find
strated age-adjusted relative risks of mor- that the measure of effect increases with
tality for persons with low social network increasing exposure levels. If inter-
indexes which ranged from 1.5 in persons mediate values of exposure had higher
with no health problems to 3.5 in disabled measures of effect than the highest values
individuals. And Blazer (32), as discussed of exposure, for example, a more complex
above, described an estimated relative relationship may exist.
risk of 3.40 (95 per cent confidence limits A number of studies have demonstrated
1.88-6.16) for 30-month mortality of el- a biologic gradient. Ware and Donald (11)
derly individuals with poor perceived so- noted a gradual and steady increase in
cial support, adjusting for 10 important positive well-being with increasing num-
potential confounders recorded at the base- bers of close friends and relatives in a
line of entry into his study. community survey. Stephens et al. (7)
Consistency. By this, Hill (38) suggest- showed a similar result for five different
ed that causal inferences could be strength- morale scales and three social activity
ened by the knowledge that the findings scales; morale and activity scores in-
have been observed repeatedly "by differ- creased consistently across five levels of
ent persons, in different places, circum- an informal social support index. In a pro-
stances and times." The studies reviewed spective cohort study, Medalie et al. (41)
have found the association of social sup- showed consistently increasing age-
port and mental or physical health in a adjusted angina pectoris incidence rates
number of groups of people in different with increasing severity scores for five
situations: community surveys (5, 11, 22, levels of family problems and four levels
23, 27, 30), elderly community residents of co-worker problems and "superior"
(7,14,16, 32, 39), middle-aged and elderly problems. These problem scores included
men (15, 31, 41), women at all stages of both low social support and conflict in in-
the life cycle (6, 13, 28, 37, 42), young men terpersonal relationships. Tibblin (44),
(29), parents (10), employed men (17, again in a prospective cohort study,
43), unemployed men (24), college stu- showed consistent increases in mortality
dents (9), Chinese-Americans (12), Swedish rates with decreases in scores on his social
men (31), psychiatric patients (20, 21, 26) network scale. Berkman and Syme (30)
and general practice patients (8). The ma- showed incremental increases in mortal-
jority of findings have been of a similar ity rates with decreasing numbers of so-
magnitude (and all have the same direc- cial connections for men and women
tion of effect) using measures of social across all age categories.
support as varied as the investigators (see Blazer (32) was unable to confirm a
"Specificity of outcome" below) and under consistent gradient of increasing mortal-
a variety of stressful conditions and health ity rates with progressive decreases in
outcomes. Inconsistencies are found, as we perceived social support or frequency of
526 BROADHEAD ET AL.

social interaction. He did note, however, chiatric disorder, it does not address the
that there was a "consistent but nonsig- relative importance of various degrees of
nificant pattern of increased mortality" other types of support.
with gradational decreases in "roles and Biologic plausibility. The biologic plau-
attachments," a summary measure of sibility and mechanisms of effect of stress
marital status and the number of living on physical health have been well estab-
siblings and children. Although not sig- lished (47-53). These include the "fight
nificant in his sample of 331 elderly per- or flight" response described by Cannon—
sons, this suggests a dose-response re- a sudden discharge of the sympathetic
lationship between the number of possible nervous system resulting in elevated blood
familial sources of support and mortality. pressure, increased cardiac output, ele-
Two studies were not able to demon- vated serum catecholamines and elevated
strate such a gradient. When controlling serum free fatty acids (49)—and "the gen-
for the presence or absence of a confidant, eral adaptation syndrome" of Selye (47)—
Miller and Ingham (8) found a relation- which includes a delayed and prolonged
ship between the number of acquain- adrenalcortical-regulated protein cata-
tances a person had and a number of bolic effect. Evidence is accumulating to
psychologic and physical symptom scores. explain a mechanism for the effects of so-
But, while persons with "few acquain- cial support.
tances" had the highest symptom scores Bovard (50, 51) has reviewed the litera-
and persons with "some acquaintances" ture supporting the existence of a brain
had the lowest scores, persons with "many system located in the anterior and lateral
acquaintances" had intermediate scores— hypothalamus which antagonizes the
an inconsistency in the proposed gradient neuroendocrine response to environmen-
This may reflect the differences between tal stress by direct inhibition of the poste-
quantity and quality of social support (see rior hypothalamic zone (responsible for
"Specificity of outcome" below) and the sympathetic activity and adrenocortico-
suggestion that the number of social con- trophic hormone release) and by its own
tacts does not necessarily reflect the num- neuroendocrine effects—parasympathetic
ber of socially supportive relationships activity resulting in lowered blood pres-
(22), that social relationships are both sure and growth hormone release, which
sources of stress and support (45) and that has its own protein anabolic effect. Auto-
an increase in the number of social acquain- stimulation of these areas with implanted
tances may produce environmental de- electrodes is reinforcing for experimental
mands for reciprocal support which exceed animals, and electrical stimulation of
the person's abilities to meet them (46). analogous regions of the unanesthetized
Brown et al. (42) tried to demonstrate a human brain have produced feelings of
change in risk of developing psychiatric pleasure, euphoria, relaxation, joy and
disorder across four levels of support: a = satisfaction. Psychotic episodes in schizo-
women with a spouse or boyfriend as con- phrenic patients have been terminated by
fidant, b = women with a confidant other such stimulation, and it has been used to
than spouse or boyfriend and seen at least control the intractable pain of terminal
once per week, c = women with a confi- cancer and rheumatoid arthritis. Bovard
dant seen less than weekly, and d = hypothesizes that any physical or social
women with no confidant. They were able stimulus which is reinforcing—such as
to detect a significant effect only for the sexual stimulation, affection, and social
highest level of support (group a). That is, approval—would act through this system
a gradient was not demonstrable. While to mediate the response to stressful stimuli.
this may reflect a threshold effect of con- Benson and others (54-56) have de-
fidant support on the development of psy- scribed an apparent counterpart to Bo-
 SOCIAL SUPPORT AND HEALTH 527
vard's "positive brain system activity" support in affecting brain or plasma /3-
which Benson calls the "relaxation re- endorphin concentrations. But Bovard
sponse." This is analogous to the relation- (50, 51) described a brain system which is
ship between the "negative brain system antagonistic to the "fight or flight" re-
activity" of the posterior hypothalamus sponse, which corresponds anatomically
and Cannon's "fight or flight" response. to sites for which repetitive electrical au-
The trance-like state of the "relaxation tostimulation in experimental settings is
response" can be self-induced by au- reinforcing and to sites which reflect in-
tohypnosis and allows individual control creased activity in the presence of rein-
over this portion of the autonomic ner- forcing stimuli (such as food, sex, and
vous system. Benson and colleagues water), and which produces major effects
(54-56) have shown its effectiveness in also associated with /3-endorphin (anal-
reversing the hypertensive effects of gesia, growth hormone release, euphoria,
stress. and addictive behavior) (50, 51, 58, 59).
Other recent work has discovered pos- Hence, it is likely that /3-endorphin is a
sible biochemical intermediates between biochemical mediator between reinforcing
sensory perception and the neuroendo- external stimuli and this response. Rein-
crine responses to stressful and socially forcing stimuli such as affection, social
supportive stimuli. These are the neuro- approval, and other forms of social support
peptides, especially the opioid /3-endorphin may also produce their euphoria, direct
(57-59). Although the exact relationship positive health effects and stress-buffering
between /3-endorphin and the positive hy- effects through /3-endorphin and/or other
pothalamic activity reviewed by Bovard neuropeptides.
(50, 51) is not known, its effects are quite Coherence. Hill (38) insisted that the
similar: analgesia sufficient to treat severe cause - and - effect interpretation should
pain, growth hormone release, and eupho- not seriously conflict with the existing
ria. Another effect of /3-endorphin is rem- knowledge of biology and the natural his-
iniscent of the "relaxation response"; that tory of disease. Indeed, social support
is, a state of narcolepsy. Indeed, Benson theory is congruous with existing knowl-
(54) recommends that the "relaxation re- edge and the mechanism of its beneficial
sponse" be induced while seated upright effect can be discussed coherently at all
to prevent falling asleep and losing its levels of biologic organization.
full benefit. Alcock (65) has reviewed how Darwin-
Plasma /3-endorphin levels have been ian forces of evolution have resulted in
shown to increase with extreme exercise the innate human needs for social contact
and to reach chronic high levels in trained including copulation and pair bonding,
athletes (60-63). It has been suggested parental care, and the sense of belonging
that this may be the mechanism whereby to a group (group bonding). These needs
running produces a state of euphoria and are so important to species survival that,
addiction (opioid effects) and has a posi- at a level much more primitive than sym-
tive effect on certain emotional disorders biotic or altruistic cooperation, their ab-
(60)—effects similar to those attributed to sence produces distress and dysfunction
"positive brain system activity." But /3- (3, 66). But low social support may not be
endorphins can only be implicated in a stressor in the sense of a stimulus which
these positive effects of running when it produces a "fight or flight" response, but
can be shown that brain tissue levels of more likely it may be that "positive brain
/3-endorphin are increased by exercise system activity" (50, 51) produced by so-
(63, 64). cial support is a necessary condition for
There has apparently been no research the growth, development, and homeostasis
to investigate the possible role of social of the organism by way of its contribution
528 BROADHEAD ET AL.

to parasympathetic activity and growth son to "form a more veridical view of the
hormone secretion. However, the threat of objective nature of self and environment"
loss of essential levels of social support (67). 3) Given an accurate perception of
may be such a stressor. external reality, one important socially
A further understanding of the role of supportive behavior may be the provision
social support in health and disease can of information about the availability of
be gained by a discussion of the possible other environmental resources, potential
mechanisms of social support effects be- changes in demands, and the probable
yond the simple concept of a basic biologic utility of planned coping strategies (69).
need for social interaction and in the con- This type of information may reduce the
text of the person-environment fit theory subjective evaluation of misfit. 4) Once a
(46). We first must consider that our re- condition of subjective misfit has oc-
sponses to the environment (be they curred, the inhibitory effect of socially
physiologic, affective, cognitive or motoric) supportive stimuli on the brain centers
are a function of both the person and the responsible for the physiology of "fight or
environment (67). The goodness of fit be- flight" may modify the effect of environ-
tween person and environment depends mental stressors by preventing a deleteri-
upon a match between the demands of the ous response. 5) Alternatively, the direct
environment and the person's abilities to beneficial neuroendocrine effects of social
meet them on the one hand, and the needs support may buffer the immediate effect
of the individual and the resources from of a "fight or flight" response in progress,
the environment available to satisfy these or 6) cumulatively prevent or reverse its
needs on the other. Although each of these adverse health consequences (49).
components has subjective and objective The majority of these plausible mecha-
values, it is our subjective person-environ- nisms are examples of effect modification
ment fit which is "the most immediate which corroborate the findings of most
antecedent" of our response to the environ- studies designed to detect interactive ef-
ment (67). The cumulative effects of our fects (5, 6, 8, 9, 14, 16, 17, 22, 28, 40, 42,
responses over time may promote health 43). It might be suggested, then, that the
and disease—both mental and physical. two studies which were unable to detect
Caplan (67) has described how social modification in the presence of direct ef-
support might protect us from the effects fects (12, 23) were a result of statistical
of a poor person-environment fit at six chance (i.e., Type II or /3 error). An alter-
levels of functioning. 1) At the most native explanation is that in their study
elementary social level, direct aid can populations the stressful life events mea-
alter objective dimensions by increasing sures used were not sensitive or specific
the environment's resources or altering enough to reflect meaningful differences
its demands. Rather than modifying the in the effect of social support between
effect of external stressors, this prevents high and low stress groups. These results
them from occurring. Also, in the pres- might have been different if an instru-
ence of potentially stressful stimuli, di- ment one step closer to the actual experi-
rect assistance (previous and/or concur- ence of daily life were used (70).
rent) can modify the effect of the envi- Experiment or intervention. There has
ronment by teaching social network been little good research to evaluate the
building skills (68), providing models of health promoting effects of social support
effective coping skills, and providing ac- intervention (33, 34). Three studies, al-
cess to coping strategies (69). 2) At the ready described, have used adequate ran-
level of perception, input from socially domized controlled designs with pre- and
supportive relationships may allow a per- post-intervention measurements of status
 SOCIAL SUPPORT AND HEALTH 529

(35-37); however the results were equiv- this definition has been operationalized
ocal in one (36), and the forms of social varies from study to study.
support and outcome measures of the Measures of social support generally
others are too different to make mean- fall into two categories—those dealing
ingful comparisons. More intervention with the quality or content of interper-
studies are certain to occur, since interven- sonal relationships and those dealing
tion is the ultimate public health goal of with quantity and other social network
all the previous studies; however, it is im- concepts such as size, frequency, density,
portant that future intervention be ade- etc. (73). In those studies which have
quately evaluated, as illustrated in a re- compared both types of measures it is ap-
cent study of self-help intervention for parent that the quality of social support is
widows by Vachon et al. (90). a stronger predictor of health outcome
Specificity of outcome. It is clear that than quantity measures (frequency of
the proposed effect of social support is not contact, number of friends) (14, 32), and
specific to any one disease state or organ quantity of social support is often not sig-
system, but ranges from the mental to the nificantly related to well-being (5, 11,
physical. This does not detract from the 13, 19).
argument for causality because the pro- Most measures of quality of social sup-
posed mechanism of effect is plausible, port have included questions pertaining
consistent with existing theory, and is de- to a confidant relationship and three re-
pendent on a complex interaction of mind ports found a beneficial effect looking at
and body mediated through neuroendo- the confidant relationship only (8, 39, 42).
crine responses that have a wide range of The dimensions of affect and affirmation
consequences (49). Here is a requirement were present in most indexes of support
of Hill's (38) that cannot be met, because quality and it can be suggested that these
in fact some exposures have multiple dis- are merely functions of a confidant re-
ease or health promoting effects and are lationship. Another variable, which may
by their nature nonspecific. Another sim- be a component of affect and affirmation,
ple example is cigarette smoking with its is reciprocity or "reciprocal affective sup-
effects as diverse as lung cancer and port." This, too, has been shown to posi-
coronary heart disease. tively affect psychologic well-being (13).
Measurement of exposure. One problem Only a few indexes have included mea-
with social support research is a lack of sures of instrumental aid (6, 7, 9,13), and,
specificity in definitions of exposure vari- unfortunately, most of these have been
ables. Various definitions of social sup- combination indexes. In one study which
port have been put forth (2, 3, 71, 72), but looked at instrumental support sepa-
the definition of Kahn and Antonucci (71) rately, it was not a significant determi-
is most comprehensive. They feel that so- nant of psychologic well-being, nor was
cial support refers to interpersonal trans- reciprocal instrumental support (13).
actions that include one or more of the One problem with a number of indexes
following: affect (expressions of liking, or scales of social support was their com-
admiration, respect, love), affirmation bination of unrelated variables. For ex-
(expressions of agreement or acknowl- ample, Ware and Donald (11), Broadhead
edgement of the appropriateness or right- (14), and Blazer (32) have shown by factor
ness of some act or statement of another analysis that the quality of social support
person), and aid (transactions in which and frequency of social interaction are
direct aid or assistance is given—includ- minimally intercorrelated and that it may
ing things, money, information, advice, be inappropriate to combine them into
time and entitlement). The degree to which summary measures. But, some investiga-
530 BROADHEAD ET AL.

tors have intentionally combined these literature. 1) McFarlane et al. (74)

constructs (12, 25), and others, whose studied family practice patients in Can-
measures are ostensibly indexes of quan- ada age 21-60 and measured size and
tity, have contaminated their data with sources of confidant-type help for six topic
a variety of possibly independent con- areas. Stephens et al. (7) studied informal
structs—perceived quality or closeness of support quality in a community survey of
interaction (6, 7, 16, 30), voluntarily ini- elderly persons in Texas. Ingersoll and
tiated social activity (11, 30, 31), hobbies Depner (75) similarly studied social net-
(31), and even marital status (indepen- works of persons over age 55 by a social
dent of marital satisfaction) (30). Simi- network analysis system which catego-
larly, the TAPPS score (an index score rizes helping individuals into three circles
representing the adaptive potential for of network membership (inner, middle,
pregnancy) used by Nuckolls and col- outer) based on their closeness and impor-
leagues (40) has been referred to by most tance to the subject.
reviewers as a measure of social support, In general, the mean network size is 9
but in addition a large number of psycho- or 10 (74, 75). The inner circle is the larg-
logic status items are summed into the est component (three to four people) and
score. the outer circle the smallest (two) (75).
Besides specificity of the content areas McFarlane et al. (74) break this down into
of social support, there is a need for more average numbers for individuals by type:
research into the relative benefits of spe- close friends (2.24 people), friends (2.21),
cific sources of support and the role situa- work-related relationships (1.43), profes-
tional contexts play in the effect of poten- sionals (1.14), spouse (0.82), other family
tially supportive relationships. For ex- (0.65), neighbors (0.17) and others (0.02).
ample, Holahan and Moos (27) found The breadth of content for six topic areas
social support in the work environment of possible discussion is widest for one's
was more beneficial for men while social spouse (5.13), close family (2.81), and
support from the family was more benefi- friends (2.49); it decreases further with
cial for women. Morrow et al. (10) found more distant relationships: work-related
that, for parents of children with cancer, individuals (1.42), professionals (1.15),
the sources of social support which were other family (0.88), and neighbors (0.19).
most beneficial varied with whether the There is important variability by sex.
child was under treatment, in remission, Women's networks are slightly larger
or had died. Wells et al. (43) found that than men's at all ages greater than 55
blue-collar workers' spouses and super- with the difference being one more indi-
visors were more effective sources of sup- vidual on an average in the "inner circle"
port than friends or relatives. Coworkers of closest relationships. Women's net-
were the least effective. works have a higher proportion of family
Determinants of social support. A de- and friends, while men's are more work-
scription of the varying distribution of so- related (74). Women also discuss more
cial support by demographic characteris- content areas with their networks and
tics points out that while much of social feel more helped by the people with whom
support may be environmentally deter- they discuss their concerns. Stephens et
mined (e.g., by social class, size of com- al. (7) found that there is no change in
munity), much of it may be determined by average amount of informal support (a
characteristics of the person (e.g., age, measure of quality and availability) by sex.
race, sex, self-initiated social activity). Marital status is also an important fac-
Three good studies of the distributions of tor. McFarlane et al. (74) found that mar-
social support have been reported in the ried individuals, of course, have more con-
 SOCIAL SUPPORT AND HEALTH 531
tact with a spouse. They also have more and access to birth control). For the el-
work-related individuals in their net- derly, social support is lowest when three
works. Single adults have a larger num- or more individuals live in a household,
ber of friends, and the widowed and di- intermediate for persons who live alone,
vorced have more professional contacts. and optimal for households of two (per-
The widowed and divorced appear to feel haps these are intact married couples).
more often that their network is not being Two personal characteristics bear spe-
adequately helpful or supportive. Stephens cial mention. The availability of informal
et al. (7) similarly found a gradient of in- social support increases with both the
formal supports with married individuals number of organizations or clubs an el-
receiving the most, followed by the never derly person belongs to and with his or
married, the widowed, and then the di- her frequency of religious service atten-
vorced. dance (7). These variables, discussed
Age (or its consequences) causes a de- below, reflect an individual's own active
crease in both network size (75) and social network building skills/abilities/
amount of informal support (7) for persons options and they emphasize the indi-
over age 55. For women the decrease vidual's role in controlling and facilitat-
seems to be steady, but most of the de- ing environmental access to social sup-
crease occurs in the "middle and outer cir- ports.
cles" with the "inner circle" stable over Dynamics of social support. Although
time. For men, the "inner circle" de- we might think of social support as the
creases gradually with age, but the most static environmental exposure status of
variability occurs in the "middle circle." an individual, it, like any exposure, may
Here, the bulk of the decrease occurs at fluctuate and has its own determinants of
about the time of retirement, and is par- variability. That is to say, social support
tially recovered by age 75 or above. is a dynamic phenomenon. It has deter-
Nevertheless, the older an individual is, minants which are internal to the indi-
the less apt he/she is to express a desire vidual (e.g., individual temperament or
for a larger network of support, but the patterns of perceiving and interacting
more likely he/she is to report health as a with the environment) and those which
restriction on access to contact with the are externally mediated (e.g., social role
network. This latter phenomenon is more definitions). Both types of determinants
prominent among females (75). are active at all stages of the life cycle.
A number of other personal character- The internal determinants of social
istics affect availability of informal social support are, in fact, coping strategies.
supports. The study by Stephens et al. (7) Coping may occur at a behavioral, cogni-
reports that, as to race whites have tive, or physiologic level (76). Although
greater access than blacks, who are more cognitive coping skills may affect percep-
well off than Mexican Americans (in tions of social support, it is the behavior of
Texas). Persons living in communities an individual in the form of social coping
smaller than 2500 in population have skills which affect the availability of so-
more informal supports than city-dwell- cial supports. A person who is either self-
ers. There are definite gradients of in- reliant or resigned to helplessness and
creasing support with increases in so- does not seek the advice of others when
cial prestige and numbers of neighbors of help is needed may have less social sup-
a subject's own age. The amount of sup- port on an average than a person who
port decreases with numbers of living copes actively by seeking advice, informa-
children (perhaps a reflection of a re- tion, or simply someone to talk to about a
lationship between socioeconomic status problem (5, 77-80). Likewise, a person's
532 BROADHEAD ET AL.

tendency to seek affiliation or social con- made by the environment, but the oppor-
tact with others even in the absence of a tunities or resources for development of a
problem will partially determine the so- social support system (71). Hence, the dis-
cial supports available (3, 69, 81). Thus, tribution of social supports varies, as we
although measures of environmental so- have discussed, with demographic and
cial support should preferably exclude in- other personal characteristics related to
dividual psychologic characteristics, a role and is not independent of life events.
person's temperament with regard to so- We see that the dynamics of social sup-
cial affiliation should be considered if an port involves a set of complex relation-
understanding of the process is to be ships between variables. Temperament
reached. and other individual psychologic charac-
Those psychologic characteristics which teristics interact with the social environ-
affect social support availability are ment during all stages of development to
under continuous change during devel- produce behavioral coping styles (or pat-
opment and have far reaching effects into terns). These in turn interact with the
adulthood. Even in early infancy we find current social environment to allow
evidence in discussions of "temperament" mobilization of social support or recruit-
for the role of the individual in altering or ment of a social network. A longitudinal
controlling the social environment— or life-span perspective is essential to an
especially with respect to the maternal- understanding of these dynamics because
infant dyad (82). Affiliative or support- the availability of social support at any
seeking skills learned in infancy are ex- one time is dependent on both the current
panded upon during childhood and ado- state of affairs (with respect to needs,
lescence "through long sequences of ex- abilities, demands, and resources) and
periences with considerable transfer of cumulative experience.
learning from one stressful episode to an- Clinical implications. What is the clini-
other" (69). The parental social network cal significance of social support? The
during childhood (68) and the peer social clinician may ask, "How will my knowl-
environment during adolescence (83) are edge of my patient's social support alter
very important in the development of pre- my treatment plan?" Knowledge of the
cursors to adult social coping skills. The dynamics of social support and its mecha-
defenses used by the eventual adult in nism of association with mental and phys-
coping, and their social consequences, are, ical health or disease results in two
as we have discussed, primary determi- categories of response—prevention and
nants of social availability (84). Thus a intervention. Prevention refers to the
person's "ability" (in person-environment suggestion by Kaplan et al. (3) of the need
fit terms) changes with age and experi- for early childhood education and adult
ence, and these abilities may have direct training in affiliation and coping skills to
effects on the "demands" and "resources" teach a person to garner the necessary so-
of the environment. cial resources to maintain his or her
In addition to the cumulative effect of health and help withstand the onslaught
experience on one's skills or abilities to of stressful life events which are the inev-
mobilize social supports, a person's roles, itable consequences of living. In the realm
needs and circumstances also change with of intervention, others have suggested
age and experience. The form and amount that physicians and other health prac-
of social support appropriate depends titioners should be trained to identify
upon these changes, and the nature of high risk families and individuals, assess
one's role after each life transition or life the nature and deficiencies of their social
event determines not only the demands networks and through "anticipatory care"
 SOCIAL SUPPORT AND HEALTH 533

(85) or "anticipatory guidance" (86) assist since the writing of Cassel (1) and Cobb
and teach them to manipulate their envi- (2), referred to the "buffering hypothesis."
ronments to acquire and maintain the so- That is, social support was conceptualized
cial supports necessary for healthy sur- as an exposure which interacted with the
vival (1, 3, 72, 86, 87). Intervention might stress of life events to reduce their delete-
also include direct mobilization of a pa- rious consequences. However, previous
tient's informal support system (family, studies of this buffering relationship are
friends, neighbors, community volun- difficult to interpret because they use in-
teers) by the clinician (88). The quality of ventories of life events which are pre-
support might also be improved by engag- cursors of change in social support, the
ing patients in mutual support groups proposed effect-modifier. Until stress and
with other patients with similar problems social support measures are "disaggre-
or situations (36). gated" this relationship will not be ade-
Conclusions and recommendations. As quately evaluated (89). There is a need for
has been suggested before (3), social sup- the development of stress measures which
port is much more than a simple envi- are not conceptually confounded with so-
ronmental exposure. It can be studied as cial support.
an effect modifier or buffer against the Other forms of effect modification or
stress of life events (1, 2), but also as a statistical interaction are often neglected.
direct determinant of health or illness (an Three are evident in the literature re-
independent variable) and as dependent viewed here, and deserve closer study. 1)
variable with its own causes and deter- Social support has an effect which is ap-
minants. A more complete understanding parently of greater benefit to women than
of social support will result from a re- men (5, 8, 27, 30). Is this a measurement
search model which considers social sup- error? Do men respond differently to
port from all three perspectives. We con- questionnaires asking social support in-
clude this review by discussing the prob- formation? Or is this a substantive differ-
lems and research needs in each of these ence? Are the social supports effective for
areas (summarized in table 2). men different than those for women? 2)
Social support as an effect modifier has, Social support may be more beneficial to

TABLE 2
Social support research needs

Social support as an effect modifier:

Need to investigate the effect of social support as a buffer against stress, using measures of stress which
are independent from social support.
Need to investigate other statistical interaction with social support including sex, previous health status,
education (and other measures of socioeconomic status).
Social support as an independent variable:
Need to address the issue of a possible third factor causing both declines in social support and health.
Need to corroborate social support theory by studies of experiments of nature as well as clinical inter-
vention.
Need to explore various definitions of social support and the relative benefits of various types and sources
of social support.
Social support as a dependent variable:
Need prospective study of personal and social characteristics which may be determinants of social support.
Need intervention studies which demonstrate effective increases in social support.
Social support and biologic plausibility:
Need to combine experimental and epidemiologic evidence on biologic mechanisms.
Need to investigate specific mediating biologic mechanisms.
534 BROADHEAD ET AL.

those already in poor health (30). Are healthy, well supported individuals is ob-
there certain groups of patients with cer- served for losses in social support and
tain types of illnesses or conditions who subsequent declines in health status. If
would be expected to benefit more from losses of social support precede declines in
social support intervention strategies health and are statistically independent
than others? 3) Education may also inter- of suspected third factors (e.g., socioeco-
act with social support so that the least nomic status), this third factor hypothesis
educated receive the greatest benefit (14). can be dismissed. However, if social sup-
Why then is the relative benefit of social port is not independent of a third factor,
support reduced for the well educated? this would not be sufficient to establish a
Could education perhaps be an indicator causal relationship between the third fac-
of verbal and other skills necessary to re- tor and social support. And, therefore,
ceive vicarious social interaction from lit- this would not disprove the hypothesis of
erature, theatre, and other arts, which a causal relationship between social sup-
nullify the effects of low social support? port and health. 2) A more clearcut line of
Or does education grant one the earning investigation might be one of experimen-
power to buy the necessary assistance and tal intervention. The argument for cau-
help, including counseling? Other poten- sality would be greatly strengthened by
tial effect modifiers need to be considered randomized controlled trials in which
as well. clinical intervention with poorly sup-
Our review of the characteristics of the ported individuals could be demonstrated
association between social supports and to produce a rise in perceived social sup-
health has concentrated on conceptualiz- port in the treatment group followed by a
ing social support as an independent lower rate of morbidity and mortality
variable, and, given this perspective, a compared with the control group. This
strong argument for a causal relationship type of study would be most efficient if
can be made (38). However, specific groups of individuals at high risk of unde-
weaknesses in this argument remain, es- sirable outcomes were used: e.g., victims
pecially with respect to a demonstration of acute myocardial infarction, post-
of the necessary antecedent-consequent operative patients, pregnant women (37,
relationship, adequate experimental or 40), automobile accident victims (19), or
interventional corroboration, and clear widows (90).
definitions and quantification of expo- The results of social support research
sure. vary with the definitions of social support
Although key studies of social support which are used. A great need, then, is an
and mortality (30—32) demonstrate that evaluation of the multiple dimensions of
poor social support precedes mortality social support to determine which have
even when controlling for baseline health significant effects on health outcome, and
status, the argument for the proposed then which have the strongest significant
cause-effect relationship is weakened by effects. Only after determining the rela-
considering that social support and health tive effectiveness of the various kinds and
may be independent consequences of a sources of support will accurate evalua-
third factor (such as socioeconomic status) tions of a patient's needs be possible, thus
with chronic effects which may limit sup- making social support a more useful clini-
portive resources and access to health cal concept.
care, thus resulting in poorer health. Finally, social support's role as a de-
Two types of "experimental" studies pendent variable has been neglected.
may help elucidate the nature of this re- Variables such as community character-
lationship. 1) An "experiment of nature" istics, socioeconomic status, social roles,
may be studied in which a cohort of social coping skills and other personal
 SOCIAL SUPPORT AND HEALTH 535
and social characteristics should be inves- 15. Graham TW, Kaplan BH, Cornoni-Huntley JC,
et al. Frequency of church attendance and blood
tigated prospectively as possible factors in pressure elevation. J Behav Med 1978;l:37-43.
the building and maintenance of socially 16. Fuller SS, Larson SB. Life events, emotional
supportive networks. It is this kind of in- support, and health of older people. Res Nurs
formation which would permit the design Health 1980;3:81-9.
17. LaRocco JM, House JS, French JRP Jr. Social
of prevention and intervention strategies support, occupational stress, and health. J
of plausible efficacy. Once designed, ade- Health Soc Behav 1980;21:202-18.
quate evaluations of these strategies are 18. Wilcox B. Social support in adjusting to marital
disruption: a network analysis. In: Gottlieb BH,
necessary to demonstrate true rises in ed. Social networks and social support. Beverly
perceived social support as well as chang- Hills, CA; Sage Publications, 1981:97-115.
es in health and outcome. 19. Porrit D. Social support in crisis: Quantity or
quality? Soc Sci Med 1979;13A:715-21.
20. Silberfeld M. Psychological symptoms and social
REFERENCES
supports. Soc Psychiatry 1978;13:11-17.
21. Henderson S, Duncan-Jones P, McAuley H, et
1. Cassel J. The contribution of the social envi- al. The patient's primary group. Br J Psychiatry
ronment to host resistance. Am J Epidemiol 1978;132:74-86.
1976;104:107-23. 22. Pearlin LI, Lieberman MA, Menaghan EG, et
2. Cobb S. Social support as a moderator of life al. The stress process. J Health Soc Behav
stress. Psychosom Med 1976;38:300-14. 1981;22:337-56.
3. Kaplan BH, Cassel JC, Gore S. Social support 23. Williams AW, Ware JE, Donald CA. A model of
and health. Med Care 1977; 15(5, suppl):47-57. mental health, life events, and social supports
4. Holmes TH, Rahe RH. The social readjustment applicable to general populations. J Health Soc
rating scale. J Psychosom Res 1967;ll:213-8. Behav 1981;22:324-36.
5. Billings AG, Moos RH. The role of coping re- 24. Gore S. The effect of social support in moderat-
sponses and social resources in attenuating the ing the health consequences of unemployment.
stress of life events. J Behav Med 1981;4:139-57. J Health Soc Behav 1978;19:157-65.
6. Barrera M. Social support in the adjustment of 25. Schaefer C, Coyne JC, Lazarus RS. The health-
pregnant adolescents: assessment issues. In: related functions of social support. J Behav Med
Gottlieb BH, ed. Social networks and social sup- 1981;4:381-406.
port. Beverly Hills, CA: Sage Publications, 26. Surtees PG. Social support, residual adversity
1981:69-96. and depressive outcome. Soc Psychiat 1980;
7. Stephens RC, Blau ZS, Oser GT, et al. Aging, 15:71-80.
social support systems, and social policy. J Ger- 27. Holahan CJ, Moos RH. Social support and
ont Soc Work 1978;l:33-45. psychological distress: a longitudinal analysis. J
8. Miller PMcC, Ingham JG. Friends, confidante, Abnormal Psychol 1981;49:365-70.
and symptoms. Soc Psychiatry 1976; 11:51-8. 28. Turner RJ. Social support as a contingency in
9. Abbey A, Abramis DJ, Caplan RD. Measuring psychological well-being. J Health Soc Behav
social support: the effects of frame of reference 1981;22:357-67.
on the relationship between social support and 29. Doehnnan SR. P-E fit theory, social support,
strain. Presented at the American Psychological and midlife career change. Presented at the
Association Conference, Los Angeles, CA, Au- American Psychological Association Confer-
gust 1981. ence, Los Angeles, CA, August 1981.
10. Morrow GR, Hoagland A, Carnrike CLM Jr. So- 30. Berkman LF, Syme SL. Social networks, host
cial support and parental adjustment to pediat- resistance and mortality: a nine-year follow-up
ric cancer. J Consult Clin Psychol 1981;49: study of Alameda County residents. Am J
763-5. Epidemiol 1979;109:186-204.
11. Ware JE, Donald CA. Social well-being: its 31. Tibblin G. Risk factors for IHD death and other
meaning and measurement. Presented at the deaths in 60 year old men in Goteburg: The Men
annual meeting of the American Public Health Born in 1913 Study. (Manuscript.)
Association, Detroit, MI, October 1980. 32. Blazer DG. Social support and mortality in an
12. Lin N, Ensel WM, Simeone RS, et al. Social sup- elderly community population. Am J Epidemiol
port, stressful life events, and illness: a model 1982;115:684-94.
and an empirical test. J Health Soc Behav 33. Gottlieb BH. Social networks and social support
1979;20:108-19. in community mental health. In: Gottlieb BH,
13. Israel BA. Social networks and psychological ed. Social support networks and social support.
well-being among elderly women. Doctoral dis- Beverly Hills, CA: Sage Publications, 1981:
sertation, School of Public Health, University of 11-42.
North Carolina at Chapel Hill, 1981. 34. Norbeck JS. Social support: a model for clinical
14. Broadhead WE. Social support, life change and research and applications. ANS 1981;3(4):43-59.
illness in elderly Durham County residents. 35. Pless IB, Satterwhite B. A measure of family
Master's thesis, School of Public Health, Uni- functioning and its application. Soc Sci Med
versity of North Carolina at Chapel Hill, 1980 1973;7:613-21.
536 BROADHEAD ET AL.

36. Gottlieb BH. Preventive interventions involv- 55. Peters RK, Benson H, Porter D. Daily relaxa-
ing social networks and social support. In: tion response breaks in a working population: I.
Gottlieb BH, ed. Social networks and social sup- Effects on self reported measures of health, per-
port. Beverly Hills, CA: Sage Publications, formance, and well-being. Am J Public Health
1981:201-32. 1977;67:946-52.
37. Sosa R, Kennell J, Klaus M, et al. The effect of a 56. Peters RK, Benson H, Peters JM. Daily relaxa-
supportive companion on perinatal problems, tion response breaks in a working population:
length of labor, and mother-infant interaction. II. Effects on blood pressure. Am J Public
N Engl J Med 1980;303:597-600. Health 1977;67:954-9.
38. Hill AB. The environment and disease: associa- 57. Guillemin R. Endorphins, brain peptides that
tion or causation? Proc Roy Soc Med 1965;68: act like opiates. (Editorial.) N Engl J Med
295-300. 1977;296:226-8.
39. Lowenthal MF, Haven C. Interaction and adap- 58. Krieger DT, Martin JB. Brain peptides. (Part I.)
tation: intimacy as a critical variable. Am Sociol N Engl J Med 1981;304:876-85.
Rev 1968;33:20-30. 59. Krieger DT, Martin JB. Brain peptides.(Part II.)
40. Nuckolls KB, Cassel J, Kaplan BH. Psychoso- N Engl J Med 1981;304:944-51.
cial assets, life crisis and the prognosis of preg- 60. Bortz WM, Angiven P, Mefford IN, et al. Cate-
nancy. Am J Epidemiol 1972;95:431-41. cholamines, dopamine, and endorphin levels
41. Medalie JH, Snyder M, Groen JJ, et al. Angina during extreme exercise. (Letter.) N Engl J Med
pectoris among 10,000 men: 5 year incidence 1981;305:466-7.
and univariate analysis. Am J Med 1973;55: 61. Can- DB, Bullen BA, Skrinar GS, et al. Physical
583-94. conditioning facilitates the exercise-induced se-
42. Brown GW, Bhrolchain MN, Harris T. Social cretion of beta-endorphin and beta-lipotropin in
class and psychiatric disturbance among women women. N Engl J Med 1981;305:560-3.
in an urban population. Sociology 1975;9: 62. Gambert SR, Hagen TC, Garthwaite TL, et al.
225-54. Exercise and the endogenous opioids. (Letter.) N
43. Wells JA, McMichael AJ, House JS, et al. Ef- Engl J Med 1981;305:1590-l.
fects of social support on the relationship be- 63. Hawley LM, Butterfield GE. Exercise and the
tween occupational stress and health. (Manu- endogenous opioids. (Letter.) N Engl J Med
script.) 1981;305:1591.
44. Tibblin G. Social factors and mortality in a pro- 64. Carr DB, Bullen BA, Skrinar GS, et al. Exercise
spective study of men aged 60: The Men Born in and the endogenous opioids. (Letter.) N Engl J
1913 Study. (Manuscript) Med 1981;305:1592.
45. Eckenrode J, Gore S. Stressful life events and 65. Alcock J. Animal behavior: an evolutionary
social supports: the significance of context. In: approach. Sunderland, MA: Sinauer Associates
Gottlieb BH, ed. Social networks and social sup- 1976:466-505.
port. Beverly Hills, CA: Sage Publications, 66. Harlow HF, Harlow MK, Suomi SJ. From
1981:43-68. thought to therapy: lessions from a primate lab-
46. French JRP, Rodgers W, Cobb S. Adjustments oratory. Am Sci 1971;59:538-49.
as person-environment fit. In: Coelho GV, 67. Caplan RD. How social support prevents social
Hamburg DA, Adams JE, eds. Coping and adap- problems in life transitions. Presented at the
tation. New York, NY: Basic Books, 1974. American Psychological Association Confer-
47. Selye HS. Stress and aging. J Am Geriatr Soc ence, Los Angeles, CA, August 1981.
1970;18:669-80. 68. Cochran MM, Brassard JA. Child development
48. Weiner H. Psychobiology and human disease. and personal social networks. Child Dev
New York, NY: Elsevier, 1977. 1979:50:601-16.
49. Eisdorfer C, Wilkie F. Stress, disease, aging and 69. Hamburg DA, Hamburg BA. Adaptation and
behavior. In: Birren JE, Schaie KW, eds. Hand- health: a life-span perspective. In: Kaplan BH,
book of the psychology of aging. New York, NY: Ibrahim MA, eds. Family medicine and suppor-
Van Nostrand Reinhold, 1977:251-75. tive interventions: an epidemiological approach.
50. Bovard EW. A concept of hypothalamic func- Chapel Hill, NC: Institute for Research in Social
tioning. Perspect Biol Med 1961;5:52-60. Science, 1981:41-55.
51. Bovard EW. The balance between negative and 70. Kanner AD, Coyne JC, Schaefer C, et al. Com-
positive brain system activity. Perspect Biol parison of two modes of stress measurement:
Med 1962;6:116-27. daily hassles and uplifts versus major life
52. Monjan AA, Collector MI. Stress-induced mod- events. J Behav Med 1981;4:l-39.
ulation of the immune response. Science 71. Kahn RL, Antonucci TC. Convoys over the life
1977; 196:307-8. course: attachment, roles, and social support. In:
53. Friedman M, Rosenman R, Carroll V. Changes Baltes PB, Brim OG Jr, eds. Life-span develop-
in serum cholesterol and blood clotting time in ment and behavior. Vol. 3. New York, NY:
men subjected to cyclic variation in occupa- Academic Press, 1980.
tional stress. Circulation 1958;17:852-61. 72. Antonovsky A. Health, stress and coping. San
54. Benson H. Systemic hypertension and the re- Francisco, CA: Josey-Bass Publishers, 1979.
laxation response. N Engl J Med 1977;296: 73. Donald CA, Ware JE, Brook RH, et al. Concep-
1152-6. tualization and measurement of health for
 SOCIAL SUPPORT AND HEALTH 537
adults in the Health Insurance Study: Social 82. Antonucci T. Attachment: a life-span concept
health. Vol. IV. R-1987/4-HEW, Santa Monica, Hum Dev 1976;19:136-42.
CA: Rand Corporation, 1978. 83. Newman BM. Coping and adaptation in adoles-
74. McFarlane AH, Neale KA, Norman G, et al. cence. Hum Dev 1979,22:255-62.
Methodological issues in developing a scale to 84. Vaillant GE. Health consequences of adaptation
measure social support. Schizophr Bull 1981;7: to life. Am J Med 1979;67:732-4.
90-100. 85. Pridham KF, Hansen MF, Conrad HH. An-
75. Ingersoll B, Depner C. Support networks of ticipatory care as problem solving in family
middle-aged and older adults. Presented at the medicine and nursing. J Fam Pract 1977;4:
meeting of the American Psychological Associa- 1077-81.
tion, Montreal, Canada, September 1980. 86. Medalie JH. The family life cycle and its impli-
76. Linden W, Feuerstein M. Essential hyperten- cations for family practice. J Fam Pract
sion and social coping behavior. J Human Stress 1979;9:47-56.
1981;7(l):28-34. 87. Dimsdale JE, Eckenrode J, Haggerty RJ, et al.
77. Folkman S, Lazarus RS. An analysis of coping The role of social supports in medical care. Soc
in a middle-aged community sample. J Health Psychiatry 1979;14:175-80.
Soc Behav 1980;21:219-39. 88. Froland C, Pancoast DL, Chapman NJ, et al.
78. Pearlin LI, Schooler C. The structure of coping. Linking formal and informal support systems.
J Health Soc Behav 1978;19:2-21. In: Gottlieb BH, ed. Social networks and social
79. Efeld FW Jr. Coping styles of Chicago adults: support. Beverly Hills, CA: Sage Publications,
description. J Human Stress 1980;6(2):2-10. 1981:259-75.
80. Mitchell RE, Trickett EJ. Task force report: so- 89. Gore S. Stress-buffering functions of social sup-
cial networks as mediators of social support—an port: an appraisal and clarification of research
analysis of the effects and determinants of social methods. In: Dohrenwend BS, Dohrenwend BP,
networks. Community Ment Health J 1980; eds. Stressful life events and their contexts.
16(l):27-44. New York, NY: Prodist, 1981:202-22.
81. Murawski BJ, Penman D, Schmitt M. Social 90. Vachon MLS, Lyall WAL, Rogers J, et al. A con-
support in health and illness: the concept and its trolled study of self-help intervention for
measurement. Cane Nurs 1978:365-71. widows. Am J Psychiatry 1980;ll:138O-4.