The Relationship Between Social Support and Physiological Processes: a Review with Emphasis on Underlying Mechanisms and Implications for Health

The Relationship Between Social Support and Physiological Processes: A Review With Emphasis on Underlying Mechanisms and Implications for Health

Bert N. Uchino John T. Cacioppo and Janice K. Kiecolt-Glaser University of Utah Ohio State University

In this review, the authors examine the evidence linking social support to physiological processes and characterize the potential mechanisms responsible for these covariations. A review of 81 studies revealed that social support was reliably related to beneficial effects on aspects of the cardiovascular, endocrine, and immune systems. An analysis of potential mechanisms underlying these associations revealed that (a) potential health-related behaviors do not appear to be responsible for these associations; (b) stress-bufferingeffects operate in some studies; (c) familial sources of support may be important; and (d) emotional support appears to be at least I important dimension of social support. Recommendations and directions for future research include the importance of conceptualizing social support as a multidimensional construct, examination of potential mechanisms across levels of analyses, and attention to the physiologicalprocess of interest.

Social relationships are a ubiquitous part of life, serving impor- tive aspects of social relationships (i.e., social support) to phys- tant social, psychological, and behavioral functions across the life- iological processes. We characterize these associations by exam- span. More important, both the quantity and quality of social re- ining the influence of social support on aspects of the cardiovas- lationships have been reliably related to morbidity and mortality cular, endocrine, and immune systems. The literature search (see reviews by Blazer, 1982; Broadhead et al., 1983; Cassell, 1976; was conducted using the ancestry approach and with PsycLIT Cobb, 1976; S. Cohen & Syme, 1985; and House, Landis, & Um- (1974-1995) and Medline (1983-1995) by crossing the key- berson, 1988). For instance, House et al. reviewed evidence from 6 words social support, social networks, or social integration with large prospective studies indicating that mortality is higher among cardiovascular, blood pressure, endocrine, or immune. Only more socially isolated individuals. These associations hold even studies whose researchers directly examined the association be- after inclusion of standard control variables such as age and initial tween social support and physiological function were included health status. Indeed, House et at. summarized evidence showing in this review. Based on this research, we examined potential that the association between social relationships and health is com- mechanisms responsible for the associations between social parable with standard risk factors, includingsmoking, blood pres- support and physiological function (S. Cohen, 1988). sure, and physical activity. We summarize the research examiningsocial support and phys- An important issue concerns the potential mechanisms re- iological processes by using both qualitative and meta-analytic sponsible for the epidemiologicai links between social relation- procedures. Major details regarding studies (e.g., type of support ships and such long-term health consequences (S. Cohen, 1988; assessment and main findings) were first characterized and ana- S. Cohen & Wills, 1985; Kiecolt-Glaser & Glaser, 1989 ). I n the lyzed in tabular form. Based on this qualitative analysis, meta- present review, we first examine the evidence linking the posi- analytic procedures were used primarily when (a) the pattern of results were equivocal and (b) there were a sufficient number of relatively homogeneous studies (e.g., similar paradigms) to reli- Bert N. Uchino, Department of Psychology and Health Psychology ably characterize the effects of interest. In addition, meta-analytic Program, University of Utah; John T. Cacioppo, Department of Psy- procedures were used to test specific hypotheses from our qualita- chology and Brain, Behavior, Immunity, and Health Program, Ohio tive analyses. The meta-analysis was performed using a commer- State University; Janice K. Kiecolt-Glaser, Department of Psychiatry cially available software package (Mullin, 1989) that provided de- and Brain, Behavior, Immunity, and Health Program, Ohio State tailed results regarding combined tests of significance levels, effect University. sizes, tests of variability regarding significance levels and effect We thank Timothy Smith for his comments on a draft of this article. sizes, and a fail-safe number. 1 Results of the unweighted meta- This study was partially supported by Grants T32-MHI8831, analysis are reported, but analyses weighted by sample size were MH44660, and MH42096 from the National Institute of Mental also performed and produced comparable results, To reduce the Health, a supplement to MH42096 from the Office of Women's Health, Grant DBS9211483 from the National Science Foundation, and the John D. and Catherine T. MacArthur Foundation. The fail-safe number represents the number of unpublished null Correspondence concerning this article should be addressed to Bert studies that would be needed to overturn the conclusions found in the N. Uchino, Department of Psychology, 502 Social-Behavioral Sciences meta-analysis. Although there is no standard fail-safe number, Rosen- Building, University of Utah, Salt Lake City, Utah 84112. Electronic thai (1984) suggests that 5k + 10, where k represents the number of mail may be sent via lnternet to [email protected]. retrieved studies, represents a reasonable tolerance level.

488 SOCIAL SUPPORT AND PHYSIOLOGY 489 problem of nonindependence for studies with multiple assess- mune systems as potential physiological pathways by which so- ments of social support, results were first transformed within a cial support influences physical health. study to z scores, averaged, and then entered into the meta-analysis. Therefore, as recommended by Rosenthal (1984), only one Correlational Studies Examining the Association statistic was included from each study. Finally, when results were Bet ween Social Support and Cardiovascular Function reported as nonsignificant, a conservative significance level of.50 was used (Mullin, 1989). Of the 81 studies whose researchers examined social support One important source of heterogeneity in the literature on so- and physiological processes, 57 focused on aspects of cardiovas- cial support and health revolves around the conceptualization and cular function. This emphasis is understandable considering measurement of support (Barrera, 1986; S. Cohen & Wills, 1985; that cardiovascular disorders are still the leading cause of death Heitzmann & Kaplan, 1988; Orth-Gomer & Unden, 1987; Tardy, in the United States and that social support has been linked to 1985; Winemiller, Mitchell, Sutliff, & Cline, 1993). In the present lower coronary heart disease (CHD) rates (House et al., 1988). review, we include diverse studies with both structural (e.g., social Conceptually, an examination of the relationship between so- network) and functional (e.g., emotional support) measures of social support and the cardiovascular system is important because cial support. Structural measures of support assess the existence of its implications for both the development and maintenance of and interconnection between various social relationships (e.g., CHD. For instance, the prognostic value of tonic arterial blood number of siblings), whereas functional measures of support as- pressure in predicting cardiovascular disorders is widely ac- sess the particular functions that social relationships may serve cepted (J. J. Smith & Kampine, 1990). Additionally, the reac- (e.g., providing emotional or informational support). The un- tivity hypothesis suggests that increased cardiovascular reactiv- derlying theme of these assessments is that they conceptually mea- ity to stress may be an important factor in the development of sure the potentially positive aspects of social relationships. This cardiovascular disorders (see Krantz & Manuck, 1984; Ma- diversity reflects, in part, the interest that social support has gener- nuck, 1994; and Matthews et al., 1986). ated in different areas of inquiry (e.g., sociology, psychology, and Because of the relatively large number of studies examining health). When the number of studies permitted it, we performed cardiovascular parameters, we now briefly review basic princi- focused comparisons between structural and functional measures ples of cardiovascular physiology. The cardiovascular system is of support to examine if they are associated with quantitatively involved in the transport of oxygen and the removal of carbon different effects on physiological function (S. Cohen & Wills, dioxide, a critical function for every cell and organ in the body 1985). In addition, when appropriate we discuss the implications (see Larsen, Schneiderman, & Pasin, 1986; and J. J. Smith & of both measures in research regarding social support, physiologi- Kampine, 1990, for detailed reviews). The heart muscle gener- cal processes, and health. ates the necessary force for the circulatory process. The vasculature (i.e., arteries, veins, and capillaries) serves as the vehicle for the pumping of the heart. Social Support and Physiological Processes The most commonly used cardiovascular measures in this review include heart rate, systolic blood pressure (SBP), and dia- More than 18 years have passed since the seminal reviews by stolic blood pressure (DBP). Heart rate, a measure of cardiac Cassell (1976) and Cobb (1976) on the importance of social chronotropy, is usually expressed in beats per minute. It is relationships for health. These 2 reviews in particular have been jointly determined by the sympathetic and parasympathetic responsible for generating interest in social support and its rela- nervous systems: Sympathetic activation increases heart rate, tionship to psychological and physical well-being. Cobb focused whereas parasympathetic activation decreases heart rate. primarily on the stress-buffering effects of social support and SBP and DBP are measures of the force of blood against the emphasized the informational value of social support processes arterial walls and are a function of both cardiac output and the (e.g., that one is cared for and loved) in fostering coping and relative state of the vasculature. Because of the importance of adaptation. Similarly, Cassell viewed social relationships as po- blood pressure in the transport of blood, it is normally a regu- tentially buffering the individual from life stressors but further lated endpoint. SBP is associated with ventricular contraction emphasized the importance of physiological processes in medi- (i.e., systole) and therefore corresponds to the peak arterial ating the effects of social relationships: pressure. DBP is associated with ventricular relaxation (i.e., diastole) and corresponds to the lowest arterial pressure. The psychosocial processes thus can be envisaged as enhancing susceptibility to disease. The clinical manifestations of this enhanced For purposes of this review, it is important to distinguish be- susceptibility will not be a function of the particular psychosocial tween tonic and phasic components of cardiovascular activity stressor, but of the physicochemical or microbiologic disease agents (Cacioppo, Berntson, & Andersen, 1991 ). Tonic or basal levels harbored by the organism or to which the organism is exposed. of cardiovascular activity provide information on the tonic (Cassell, 1976, p. 109) physiologic state of an individual. The correlational studies examining the association between social support and cardiovas- As suggested by Cassell, the associations between social support cular function have focused primarily on tonic measures. The and physical health have been found on such diverse heath out- phasic or reactivity components of cardiovascular activity refer comes (e.g., coronary heart disease, cancer, and infectious to momentary fluctuations from tonic levels. Recent laboratory illnesses) that there are probably multiple physiological path- studies, reviewed later, have focused on the possibility that so- ways by which social support may influence disease states. In cial support may reduce cardiovascular reactivity to acute psy- this review, we focus on the cardiovascular, endocrine, and ira- chosocial stressors. 490 UCHINO, CACIOPPO, AND KIECOLT-GLASER

An important issue to consider is the psychometric properties derlying the relationships between social support and cardiovas- of the physiological assessments because they bear on the po- cular function. tential mechanisms linking social support to long-term physical Many of the studies on social support and cardiovascular health. In this regard, it is conceptually important to distinguish function have used a correlational design with normotensive in- between measurement reliability and temporal stability. Mea- dividuals. Table l summarizes 28 correlational studies, most surement reliability refers to the accurate assessment of the of which used middle-aged and older adult samples from the physiological state at one point in time. In comparison, tempo- community. Twenty studies examined both men and women, 5 ral stability refers to a dispositional characterization of physio- examined only men, 2 examined only women, and l study did logical function (i.e., stability of the physiological assessment not report the gender composition of the sample. Researchers across different situations and occasions). Adequate measure- of 14 of these studies explicitly assessed some aspect of familial ment reliability is necessary but not sufficient for temporal sta- support. In addition, researchers of 7 studies assessed structural bility. The distinction between measurement reliability and measures of support, of 15 studies assessed functional measures temporal stability is important because if social support is to of support, and of 6 studies assessed both structural and func- have effects on disease processes with a long-term etiology, the tional measures of support. physiological assessments should be characterized by temporal In general, the results of the correlational studies are consis- stability. The assessment context (e.g., specific tasks), popula- tent with the notion that higher social support is associated with tion (e.g., phobics), and techniques (e.g., specificity of tracers better cardiovascular regulation (e.g., lower blood pressure). In in radioimmunoassay) may all influence an individual differ- 1 of the first studies investigating the relationship between social ence assessment of physiological function. As an example, a support and cardiovascular function, Kasl and Cobb (1980) ex- needle stick is often associated with relatively short-term eleva- amined the influence of social support on blood pressure tions in catecholamines. Because of the measurement reliability changes in response to job termination; they reported that per- of current techniques (Baum & Grunberg, 1995), the catechol- ceptions of social support were negatively related to blood pres- amine changes due to venipuncture would be accurately as- sure changes in response to job loss. To summarize Table 1, sessed at that point in time. However, this may be a poor index researchers of 23 studies reported some evidence that social of an individual's cathecholamine response across time and support was associated with better cardiovascular function, of situations. 4 studies reported no relationship (see Ely & Mostardi, 1986; Past researchers have examined the temporal stability of heart Houben, Diedriks, Kant, & Notermans, 1990; Kaufmann & rate, SBP, and DBP reactivity. As reviewed by Manuck, Kaspro- Beehr, 1986; and Lercher, Hortnagl, & Kofler, 1993), and of 1 wicz, Monroe, Larkin, and Kaplan (1989), measures of heart rate reported opposite effects (Hansell, 1985). A meta-analysis of reactivity evidence the strongest test-retest correlation, typically 21 correlational studies whose researchers reported data on the ranging from .67 to .91. SBP reactivity tends to evidence adequate association between social support and blood pressure revealed a significant combined test (z = 4.22, p = .00001, fail-safe n = temporal stability that is slightly lower than the stability seen for 117.38).2 The mean effect size (r) was .08, suggesting a small heart rate, whereas DBP tends to show relatively low test-retest but reliable effect across studies. None of the tests of variability stability. Although many of the studies reviewed by Manuck et at. was significant (p > .45). Thus, the evidence for an association did not report data on the test-retest stability of tonic measures, between social support and lower blood pressure levels appears the patterns of stability across heart rate, SBE and DBP appear reliable. similar to that for reactivity assessments. It should be noted, how- We coded each of the studies included in the meta-analysis as ever, that researchers have demonstrated that the stability of these measuring structural or functional measures of support. Of the cardiovascular assessments, including DBE are enhanced consid- 6 studies that assessed both types of support, we were able to erably when assessments are aggregated across multiple time separate the effects in 4 of these studies. Therefore, data from 9 points and multiple tasks (Kamarck, ! 992; Kamarck et al., 1992; studies were identified as structural, and data from 14 studies Manuck, 1994). There are several methodological issues related to an examination of the relationship between social support and cardio- 2 The meta-analysis consisted of 21 studies that directly examined the vascular function. In particular, the use of appropriate statisti- association between social support and tonic blood pressure levels. In 2 cases, we averaged the results reported across 2 different published stud- cal controls is important as many of the studies reviewed in this ies (i.e., Dressier, 1980, 1983; Janes, 1990; Janes & Pawson, 1986) be- section are correlational studies in which potential associations cause data were apparently reported on the same sample. In addition, with confounding variables may occur. For instance, social sup- the 5 studies examining job-related social support were excluded from port may be correlated with socioeconomic status, medication this analysis because it had been identified a priori as a feature associ- use, age, and other factors that may have direct influences on ated with inconsistent effects. In the text, we examine in detail potential physiological function. We should note that there is some dis- reasons why job-related support may be associated with weak effects on crepancy in the literature on whether such variables are poten- blood pressure. tial confounding variables or mechanisms by which social sup- In our initial search, we excluded 2 studies examining the relationship between social support and blood pressure for methodological reasons port has an association with health (S. Cohen, 1988; House et (James, LaCroix, Kleinbaum, & Strogatz, 1984; Orth-Gomer, Rosen- al., 1988 ). In our tabular analyses of each study, we explicitly gren, & Wilhelmsen, 1993). More specifically, these studies included note when such statistical controls were used. In addition, we participants on cardiovascular medication but did not account for this discuss the attention (or lack thereof) paid to appropriate sta- factor in reporting the association between social support and blood tistical controls and its implications for the mechanisms un- pressure. SOCIAL SUPPORT AND PHYSIOLOGY 491 were identified as functional measures of support. Focused Jones, Scott, & Adcock, 1980). Therefore, these data suggest comparisons between the structural and functional measures of that social support is important for both men and women but support revealed no differences in significance level (p = .29) or that specific types of social support may be important as a func- effect size (p = .47). Although appropriate caution is warranted tion of gender (e.g., Bland et al., 1991 ). because of the small number of studies contrasted, these data A noteworthy feature of Table 1 is the 8 studies whose research- are consistent with the larger literature, suggesting that both ers examined the association between social support and cardio- structural and functional measures of support predict benefi- vascular function in differing cultural contexts. For instance, re- cial effects on physical health. However, the specific psychologi- search by Dressier and colleagues suggests that culturally impor- cal and behavioral mechanisms that contribute to these effects tant aspects of social support predict lower blood pressure levels may differ for structural and functional measures (S. Cohen, (Dressle~ 1980, 1983, 1991; Dressier, Mata, et al., 1986; Dresslei; 1988). We return to a discussion of such issues later in the Santos, & Viteri, 1986; Dressier et al~, 1992). In addition, Janes review. and colleagues (Janes, 1990; Janes & Pawson, 1986) have data Researchers in 3 of the 4 studies that did not find any rela- indicating similar associations in Samoan culture. An important tionship between indices of social support and blood pressure point of the data generated by these studies is an emphasis on the regulation measured job-related social support (Houben et al., proper operationalization of social relationships within a specific 1990; Kaufmann & Beehr, 1986; Lercher et al., 1993). How- cultural context. For instance, Dressier (1983) used historical- ever, Winnubst, Marcelissen, & Kleber (1982) and Unden, cultural analyses to argue that societal membership is an impor- Orth-Gomer, & Elofssen (1991) also examined work-related tant facet of social relationships in West Indian culture. These social support and reported some effects on cardiovascular cross-cultural data demonstrate the generality of the effects of so- function. One potential reason for this discrepancy may be re- cial support in predicting blood pressure across differing cultural lated to the psychometric properties of the measures of social contexts. support. For example, the Lercher et al. measure of social sup- Despite the consistency of the associations reported in Table 1, port was two dichotomous questions (also see Houben et al., there are several important issues relating to these studies. Many 1990, which also contains two questions), whereas Winnubst et of the studies in Table 1 have conceptualized social as a unidimen- al's measure contained five questions and Unden et al's mea- sional construct and operationalized social support in terms of sure contained six questions. Although Kaufmann and Beehr general levels of social integration or perceptions of support. did report high internal consistencies for their job-related social Clearly, social support can be conceptualized as a multidimen- support measures (. 59 < internal consistencies < .88), they did sional construct (S. Cohen & McKay, 1984; Cutrona & Russell, not report data on the main effects of social support on blood 1990), and specific dimensions of social support may be more pressure (only that the interaction between social support and effective when they meet the demands of related stressors job stress was nonsignificant). (Curtrona & Russell, 1990). In addition, an examination of spe- It is also possible that job-related social support may not be cific dimensions of social support may suggest more precise mech- related to blood pressure because these relationship are not as anisms through which social support influences health (Uchino, significant for the individual, at least compared with other Cacioppo, Malarkey, Glaser, & Kiecolt-Glaser, 1995). To this sources of support. Consistent with this possibility, we review point, researchers in only 4 studies of Table 1 have assessed evidence later in this article indicating that familial sources of multiple functional dimensions of social support (Hanson, Isacs- support appear to be associated with reliable effects on blood son, Janzon, LindeL1, & Rastam, 1988; Kaufmann & Beehr, 1986; pressure regulation. Knox, Theorell, Svensson, & Waller, 1985; Strogatz & James, An issue of interest concerns potential gender differences in 1986). In l study that examined two conceptually distinct func- social support processes (Shumaker & Hill, 1991 ). In this re- tional aspects of social support, Strogatz and James found that gard, the studies in Table 1 generally suggest that social support lower tangible but not appraisal support was a significant predictor predicts better cardiovascular regulation in both men and of increased hypertension (i.e., DBP > 90 mm/Hg, or current women. Many of the studies in Table 1 had the effects of gender use of hypertensive medication), particularly in low income Black statistically controlled. However, 8 out of the 20 studies whose participants. These data are consistent with the notion that specific researchers examined both men and women did report gender support components may be more effective when they meet the effects (Bland, Krogh, Winkelstein, & Trevisan, 1991; Cotting-' demands of related situations. ton, Brock, House, & Hawthorne, 1985; Dressier, Grell, Gal- lagher, & Viteri, 1992; Dressier, Mata, Chavez, Viteri, & Gal- One concern in these correlational studies was the rarity in lagher, 1986; Jackson & Adams-Campbell, 1994; Janes, 1990; which psychometric data regarding the measurement of social Linden, Chambers, Maurice, & Lenz, 1993; Livingston, Le- support were reported. Only I 1 studies made reference to the vine, & Moore, 1991 ). In an illustrative study, Janes examined psychometric properties of their scale (e.g., factor analysis and both structural (i.e., social resources) and functional (i.e., internal consistency). Given the heterogeneity in which studies instrumental) measures of support in Samoan men and summarized in Table 1 have conceptualized and measured so- women. Results revealed that both structural and functional as- cial support, the scales' psychometric properties are important pects of support predicted lower blood pressure in men and to examine, especially for the less validated measures of women. However, subsequent analyses by gender revealed that support. social resources were a stronger predictor of blood pressure in Relatedly, only researchers of 4 studies in Table I reported men, whereas instrumental support was a stronger predictor of any data on the temporal stability of the cardiovascular assess- blood pressure in women (also see Henderson, Byrne, Duncan- (text continues on page 499) 492 UCHINO, CACIOPPO, AND K1ECOLT-GLASER

= t,.. "O

v 0~.. I~ "0 ='0

-= ~ ~ ~ .- .~ ~ .~ 0

.__..= ~ •= .=_ ,. =~,0 ~ ~.~ ~.~ ~ _ ._ -.- ~=., = ~ ~ ~: ~-~ -~ =~ _~=~_..~o=O_=~ gg~=

=.~=- N o ~.__, ~ 5 ~ . = . u, ,,., = o

m ~ ¢" -- '7 ~ 0 0

o~E 0 N ~ 8 .=. =

,EE

E 0

o= =~,~ ~ ~ ~ e,, -=o 0 __m

rn mf oo

'~ =o,, ~ =- SOCIAL SUPPORT AND PHYSIOLOGY 493

(b, ~ 'J

.~ ~ ~ ~'~

~ ~'~ ~ ~

,. 8. =-

Z z < Z < Z

r~ ~".~ ~,,~

__. ~-~.~ ~ ee) {') "~0 ~.., J~ $.,, "o~ ~

r-.e"~ ~ II t.,q c',I ~ ~)

.-~

0 0 0 0

~ oe

~0 v~ = 0 494 UCHINO, CACIOPPO, AND KIECOLT-GLASER

~ ~z~ ~., "~ ~'~.~ ~.. ~ .~ ~-~ ~ ~~'~ z

o 6 m ~.~.~'=".~," ~'=~'~~m e-

='~ ~, =~.~ ~#v o~,~= ~_

~ ~=~_ __. k~ ~ ~ #,~~4~ ~s o.~ ~ o~

•=.oo° 8 ,o°°°..°,~o ~.~ ~..=.~ ,o .....~ ~.~. 8 ~..o.,,..~ = ~

• ~^ ~ ~'~

O -~'~

~ ~_ e~ooo ~ .., ,, ~e'n 6"~.~

NI N

~2~ ~_'~'~ ~

g O N N i_.

e~ SOCIAL SUPPORT AND PHYSIOLOGY 495

~" ~'~'c, = ~ .~.g ~'~ ~ ~.~ ~ ~ ~ o ~ ° ~:~

~. ~ .~~ ~oE ~ ~.~ 3 ~.~ ~~ ~ .~~.~

~8 2 z .J 0

.~ ~ E ~ '.~ Q. r~ d~ ~Q ~o

= ~"~ ~'a ~'~ O~

~.- .~__ ~.~ ~-~ ,.~u. ~,~ ~.~ ~.~ ~-~ e. ;~, -~

E ~.~ 0 <0.., ,,~

{"4 f-I

e. 0 .o o 0

0 0 0 (Jo

E 496 UCHINO, CACIOPPO, AND KIECOLT-GLASER

~o "0

'~- =~.E = ,,~ o =,-.,= ,-- =

~'.= ,,', = ._ o ~ .=

e-'-- ~

r.~

ge! oE

~O ~ ~ ~I-4 ~ N ~ ~ ~.~ ~=~ 'E ~ = 0 t.,,

N,~ 7 0

e~ e~

E .-d..£ .~ ~

Eo~u

.... .,.£ ~ 'o - ~.~

- ~ ~x~'~ ~ ~--~ ~ ~ 2 ..5=

f'-l~

1¢3 -43

"N "s t- 0 0 O

0 0 (JO

v e~ .8 "O 0 e- SOCIAL SUPPORT AND PHYSIOLOGY 497

~.~ ~.~

0 ~ ... ~, i "~~ i~.~- ~-~ i.~

~E ,~-- ~, ,oo ,- ~ i~

~.~ ~ ,~ ~u'~

~o ,~.,~ ,,~ e- o ~,~,~ ~ ...~ ~-~ ,-, .g~=

~.~ iu'u ~,~ ~1 ,,,-, ~,~, ~" ~ ~ I~ ~ "~"' ~'~ ~"" Z ~ ~

r~ a2 r~

e~ 0 ~ 6

:~,o ~.~,~ Oo,~o

~Z=-u~ K.~ ~..o o~: ~~ ~.~ ~

"~ o ~ '~ .~ =I 0 i~ ° o r~ m e~

e~ ~6 0

I ~s~Z

¢- Q ~ o ~ ~ "'.~ 0 0 0 m ",~ "~ ~ O.,~

0 0 0 (..)

O~ E~ 6 -= :E,-] 498 UCHINO, CACIOPPO, AND KIECOLT-GLASER

i. "0~ ~ ~ ~=0

~,.~ ~-i= ~ 0 ~ ~ ~ O~

_~o ; '~ =: ~ I=I .,-'~ o~.'o ~.~ ~.~ ~ ~

•~ ,~ .~ ~..~ •~ ~ 0 0

0 =_

~'~ ~ 0 ,- ..,~ ~ ~ -- --,.~ ¢~ 00~ ~ ~ ~ 00~ ~ ~ Z

c4~

~m 0 ii

c~ Ox m

.. .u "0 .E SOCIAL SUPPORT AND PHYSIOLOGY 499

ments (Dressier, 1980, 1983; Hansell, 1985; Kaufman & Beehr, 1986). As a representative study, Dressier (1980) assessed SBP and DBP three times during the study, with each assessment '-.~ ~'~ ~ 0-.,.., ~, ~ t~ ~ e~"~ ~ ~ 0 separated by 20 min. Dressier reported strong average corre- 8~< lations between these reading ( r =. 89 for SBP, r = .88 for DBP) and thus evidence for temporal stability of the tonic assessments over relatively short periods of time. Nevertheless, the temporal "~ ~ ~ e~'~ e'~.-- ~" ~ ii ..- stability of the physiological measures needs examination more o~o consistently in correlational studies if social support is to be associated with the long-term development of CHD. ~.~ Finally, an important methodological issue for the studies summarized in Table 1 concerns the lack of explicit controls for cardiovascular-altering medications (e.g., beta blockers), espe- e~ ~ ~ e~ ~ ~1;,~ e., ~ O Zil~. cially because most of these studies tested middle-aged to older ~.~ adult samples. Only 10 of the studies summarized in Table 1 directly reported that they had excluded individuals on cardiovascular medication or statistically controlled for this variable ~'~, in their analyses. Researchers of 3 studies examined a young population in which medication use may not have been prob- II e.., lematic. This issue is important because one might expect indi-

.~Z viduals low in social support to be characterized by higher blood pressure levels, however, the use of cardiovascular medi- t~ ¢" o ~ cations would serve to lower their blood pressure and eliminate or weaken the association between social support and tonic blood pressure levels. In fact, this may be one factor responsible for the relatively small, albeit reliable, meta-analytic association reported in this review. Future studies need to be sensitive to O wOE this potential confound. ..- ~ H I .~_< Intervention Studies Examining the Effects of Social

,.~ e'~ o Support on Cardiovascular Function

.-~= The evidence summarized in Table 1 suggests an association < < between social support and cardiovascular function. However, 0 ~ the correlational designs may limit the inferences from these studies. Table 2 summarizes 6 prospective intervention studies with normotensive middle-aged and older adult samples that examined social relationships and cardiovascular function. The social support interventions in these studies varied: I study pro- ~ e.. 0 vided opportunities for increased social interactions in activity groups, 2 studies used group discussions to increase social interactions, 1 study contained a social support education component, and 2 studies used social support to facilitate exercise, lifestyle, and stress management training. Two studies tested only women, 3 studies both men and women, and the remaining e~ study did not report the gender composition of the sample. 0 0 Because of the small number of studies and the diverse ways in which social support was operationalized, we performed a qualitative analyses of the studies in Table 2. In this regard, researchers of 4 of the 6 studies in Table 2 reported positive effects .~. ~ ~ of social support on cardiovascular function. In a prototypic investigation, Andersson (1985) studied older adults who reported problems with loneliness and randomly assigned them to either a small group discussion (e.g., opportunities for leisure activities) that met four times during the intervention or a con- 7.o. ~ trol condition of individuals followed for the duration of the study. A 6-month follow-up revealed that the support interven- Z ~'~ tion participants had a greater frequency of social contact and 500 UCHINO, CACIOPPO, AND KIECOLT-GLASER

t,,.~ 0 ~ t'-,

ea O.

"~ .~ ,:,,~ '~ ""~ ~ &

r~ r~

= o 0 ~'~ r~ ~'~ r~

0 i£ o ~t

~. ~E ~"

0¢-I

0 0 7 r.

"4~" .8--. < o < SOCIAL SUPPORT AND PHYSIOLOGY 501

ll E O ~ "~ O •-~ ~,

= ~ ~.-~ J=

~ ~ ~ m 0 ~,~ .~ ",~.

~*., " g ~ 0 : N

•~ ~..~ '; o = , $~0 ~-~ m ~--~ :~ ~ ," ~"~ I: = I O N >° e~ = ~-~-~ ~'~ ~ o _= <

•~ .~ .~- > ~~ ~2 e~ O. rr~ r~

O ,.~ ~

~" ,,~ ~_~'~.~.~ ~2"~ =.~ ~ ~ed

~.~.o_~ ~,~ Z~ .-Z= =~ = ~ ~ ~" ~'- I _~.~~-~ ,~.~ ~.~0 ,.~.= ,~, ..,

,.= o =~°'6 6 •r-II ~, 0 ..I= O

o -r

~t'N

.. ~

, --~ 0 0 e', N E

= g 2Z:

,Ng m eg ~

-i

= 502 UCHINO, CACIOPPO, AND K1ECOLT-GLASER leisure activities and, more important, lower SBP and DBP. The issues raised by these studies. The prospective intervention control condition evidenced no such changes. studies with hypertensives were primarily designed to affect tan- Only researchers of 2 of the studies conducted to date re- gible aspects of support. However, the social support manipula- ported no effect of a social support intervention on blood pres- tions may have affected other aspects of social support, includ- sure regulation (Arnetz, Theorell, Levi, Kallner, & Eneroth, ing appraisal support due to the increased participation and 1983; Gill, Veigl, Shuster, & Notelovitz, 1984). However, Gill et knowledge of the support provider. Furthermore, although al. did not find a significant manipulation check on social sup- these studies suggest tangible support may have been important port, which suggests that the intervention was unsuccessful in because of better medical adherence (e.g., Levine et al., 1979 ), affecting participants' support networks. The null finding by none of the studies researchers performed statistical analyses to Arnetz et al. is more difficult to explain. However, familial rela- directly examine the importance of this factor. Interestingly, the tionships may be relatively important for blood pressure regu- prospective intervention studies generally preceded the correla- lation (see Potential Mechanisms). In general, stronger associa- tional studies summarized in Table 1 that suggest tangible fac- tions might be obtained with interventions that focus on famil- tors alone cannot explain the associations between social sup- ial sources of support. port and blood pressure regulation. Therefore, the prospective Although the studies summarized in Tables 1 and 2 suggest that data are only suggestive of tangible support influences on blood social support influences cardiovascular function, few of these pressure regulation because other unmeasured components of studies suggested its effect on established risk factors. To this point, social support may have contributed to these effects. we summarize 8 prospective (primarily intervention) studies with We should also note that aspects of several interventions with hypertensive patients in Table 3. Researchers of 5 of the 8 studies normotensive participants might have affected other health- in Table 3 explicitly noted that they examined both White and related processes (e.g., Andersson, 1985; Sallis, Trevorrow, Black participants. Researchers of 6 studies examined both men Johnson, Hovell, & Kaplan, 1987). For instance, the Sallis et al. and women, whereas of 2 studies examined only men. Six of the manipulation also informed participants of the harmful effects interventions used for the most part familial sources of support. of stress (also see Clifford, Tan, & Gorsuch, 1991 ). Therefore, However, researchers of I study simply assessed naturalistic social lifestyle or behavioral changes related to stress, but not directly support, and of 1 study used organizational social support. involving social support, may have also contributed to the re- Evidence for the role of social support on cardiovascular suits of these studies. Nevertheless, the prospective design of the function and risk factors comes from the prospective interven- studies in Tables 2 and 3, along with the importance of blood tion studies on hypertensive individuals summarized in Table pressure regulation in hypertensive individuals, provides rela- 3. In an early study, Levine et al. (1979) identified 400 hyper- tively strong evidence linking social support to risk factors. tensive patients and assigned them to interventions consisting of an exit interview, family support, small group, various com- Potential Mechanisms Linking Social Support to binations of these groups (e.g., exit interview and family Cardiovascular Function support), or a control condition. In the family support condition, patients were asked to identify a target individual with Because of the consistency of the associations between social whom they had frequent contact (typically a spouse). The support and cardiovascular parameters presented in Tables 1 to target individuals were then trained to increase understanding, 3, we now turn to specifying the potential mechanisms respon- support, and reinforcement regarding positive management of sible for these covariations. In a review of potential mechanisms the patient's hypertensive state. Results revealed that family linking social support to health, S. Cohen (1988) suggests that support alone decreased DBP (i.e., DBP was below the hyper- social support may have beneficial effects through social (e.g., tensive limits for the participant's particular age group) by 11% stress buffering), psychological (e.g., affective states), and be- at an 18-month follow-up assessment. Predictably, exposure to havioral (e.g., health-promoting ) mechanisms. Consistent with all intervention conditions was associated with the best blood S. Cohen (1988), we examined the mechanisms linking social pressure control (28%). Subsequent follow-ups of this project support to physiological processes at different levels of analysis sample revealed reliable long-term effects of the social support (also see Cacioppo & Berntson, 1992 ). manipulation on blood pressure regulation (Morisky, DeMuth, At a social psychological level of analysis, it appears that fa- Field-Fass, Green, & Levine, 1985; Morisky et al., 1983). A milial sources of social support may be associated with reliable meta-analysis of studies whose researchers have used social sup- effects on blood pressure regulation. A meta-analysis of 12 cor- port manipulations to control blood pressure in at-risk popula- relational studies whose researchers explicitly noted that they tions (Earp, Ory, & Strogatz, 1982; Erfurt, Foote, & Heirich, 1991; Levine et al., 1979; Morisky et al., 1983, 1985; Stahl, Kelley, Neill, Grim & Mamlin, 1984) revealed a significant 3 The meta-analysis consisted of 4 studies that used social support combined test (z = 3.32, p = .0004, fail-safe n = 12.29).3 The manipulations to control blood pressure in at-risk populations. One test statistic is entered for the 3 Levine et at. and Morisky et at. studies above mean effect size was r =. 15, and no test of variability was sig- to reduce potential problems with the nonindependent samples. The nificant (p > .20). These prospective data from at-risk popula- Pinto, Sirota, and Brown (1985) study was not included because no tions provide evidence that social support may have beneficial statistics were presented for their case study. For all prospective studies, effects on established risk factors. the statistic entered was the difference in blood pressure level or control Although the results of the prospective intervention studies between the social support manipulation and a control condition during with normotensives and hypertensives suggest that social sup- the final assessment, thereby providing evidence on the long-term effects port leads to better blood pressure regulation, there are several of these interventions. SOCIAL SUPPORT AND PHYSIOLOGY 503

.~,-- ~,~ i., I.-,"'o 0 ~..~ ~ w ~ '~~" •=,,~°''~. ~ ,~ ~.~ ,~t= E~ = o 8 ~o ~ o, Z a~'- o "~ u ~ ~ ~ ~.,-, ~- ~

¢; 0 0~. ~= ~'~'~ *-, 0 ,2--,~

'~-o ~ o~ ~.=. ~_ ^-- ~-~ 0 0 0 0 0 ~ ,<'~ E •~ ~.~ 0 Z

E~ c~

r. '~ '0

~..~.

.~.~ ,~ • . ~ ~: ~N_

0 ..-, ~ ,,-. Ca

eS~=e ~-~ 8 ~~ 8

~.~ 0~

~B .~- ~.~,

e'~ o ,~ ~', ~ _~= cq ca ~ oUu ~_o ~ .~ ,~m. ~-~,~ (~ I i ~ 0 ~

E --- ~IE ~'a ° ~'- ~0~ ~ ~ Ca ,~..d t~l I:: ~ I:: ~=u .=~ '~o .~~ .-~ .~~.~.~ ,~ > .~o'~ ,~ "~.-~'~ ~ .~ ~u.~, ~ ~*" ~= ~.~

o', o', t~

E~ i = 504 UCHINO, CACIOPPO, AND KIECOLT-GLASER

~E~ "~=~o ~:~= ~: ~ ,.~o~.~..9o o ,,~.-, ~~o,-.-. o~E~oo,~ 'E' ~ II

.: ~ ~.~'.~,.o ~t:, " >~.~ ~ 0 o ~ ,~ :n ,'~ .~ u ,~ ~ ~:n o~ •~ VI e~

..~ ~.--'~.-~ ~,~a~ ~ ~,-~ e~.~ ~ ~ ~

i:: ~ ~ ~ - ~: ~,;..~c~ 0"~ 0 0 ¢: "~ 0 0 ~ ~ 0 ~.'~ 0 0 ,,- ~:

t:Q

0 U

0 0 ~ 0 o "" .~ o ~o~

•. o~ ~ ~ ~ ~.- ~.~ 0p~ ~ 0 ' .. 0

o o

c~ .o oo e~ ~J~ 0 c~ Z < o~ ~

o.-~ .~.~ '~egg

< -- <

-Q

0 "0 ~g ~ 0 O0 0 ~ 0 (%2 w-~ •~ .~ ~ .~ .~ .. e~¢; 0 ~ ~-~ ~ 0 -0

o~ ~ m

0 o'3 z~ J~oo o Ov u~ SOCIAL SUPPORT AND PHYSIOLOGY 505 assessed social support related to family members revealed a physical activity). Contrary to this position, the associations be- reliable combined test of significance (z = 4.17, p = .00001, tween aspects of social support and cardiovascular function re- fail-safe n = 64.97). The associated effect size for this analysis mained significant even after statistically controlling for a num- was r =. 12, and no test of variability was significant (p > .09). ber of health-related variables, including weight or body mass In 1 study whose researchers directly examined blood pressure (e.g., Bland et al., 1991; Janes & Pawson, 1986; Stavig, Igra, & in the presence of different social relationships, Spitzer, Llabre, Leonard, 1984). However, it should be noted that many of these Ironson, Gellman, and Schneiderman (1992) found that being researchers have not assessed specific health-related behaviors around a family member was associated with lower ambulatory (e.g., substance abuse). In addition, of those researchers that SBP and DBP compared with being around a friend or a did assess specific health-related behaviors, data on the reliabil- stranger. The prospective interventions with hypertensive pa- ity or validity of their assessments were typically not reported tients that directly used family members as sources of support (see Umberson, 1987). provide convergent evidence on the importance of familial At a psychological level of analysis, perceptions of stress, feel- sources of support on blood pressure regulation. ings of controllability, intrusive or ruminative thinking, feelings Researchers of 8 of the correlational studies have directly of loneliness, depression, and other emotional processes (e.g., tested the potential stress-buffering effects of social support on anxiety) are potential psychological mechanisms for the associ- cardiovascular function (Dressier, 1980, 1991; Dressier, Matar, ations between social support and cardiovascular function et al., 1986; Janes, 1990; Kasl & Cobb, 1980; Kaufmann & (Collins, Dunkel-Schetter, Lobel, & Scrimshaw, 1993; Pierce, Beehr, 1986; Strogatz & James, 1986; Winnubst et al., 1982). Sarason, & Sarason, 1991; Quittner, Glueckauf, & Jackson, As argued by S. Cohen and Wills (1985), one methodological 1990; Russell & Cutrona, 1991; Stokes, 1985; Solomon, Miku- requirement for a test of the buffering model consists of demon- lincer, & Hobfoll, 1986). Unfortunately, empirical data are un- strating a significant main effect for the stress assessment to en- available concerning the psychological mechanisms responsible sure that the measure was characterized by an adequate range for the associations between social support and cardiovascular of scores and measurement reliability. However, only research- function reported in Tables 1-3. Future research is clearly ers of 4 of these studies reported data indicating that their mea- needed in this area of inquiry. We return to this important point sure of stress was associated with blood pressure. These 4 stud- later in the review. ies were associated with a significant combined test (z = 3.39, p = .0003, fail safe n = 12.99) and an effect size ofr =. 18. No test Laboratory Studies Examining the Effects of Social of variability was significant (p > .34). In 1 illustrative study, Support on Cardiovascular Function Dressier (1980) reported an interaction between structural measures of support (i.e., no. of siblings) and levels of life stress Whereas the prior studies have focused primarily on tonic for SBP and DBP: Individuals high in number of siblings and measures of cardiac function, many of the recent studies have low in life stress were characterized by the lowest blood pres- been experimental, laboratory studies conducted under the ru- sure. Although Cohen and Wills suggest that buffering effects bric of the reactivity hypothesis. Briefly, the reactivity hypothe- are more likely to be found when there is a reasonable match sis suggests thatexaggerated cardiovascular reactivity to stressors may be a pathogenic mechanism influencing the develop- between the stressor type and support function, they also reported that buffering effects were sometimes found when re- ment of cardiovascular disorders (see Krantz & Manuck, 1984; searchers assessed close interpersonal relationships. Consistent Manuck, 1994; and Matthews et al., 1986). These 15 studies with Cohen and Wills, all 5 studies whose researchers examined are summarized in Table 4. Thirteen of these studies tested rel- familial relationships (e.g., spouse and siblings) reported a sig- atively young participants under the age of 30. One study used nificant buffering effect on cardiovascular regulation (z = 3.43, a middle-aged sample, and 1 study used an older adult sample. p = .0003, fail-safe n = 16.74), with an effect size of r =. 14. No Researchers of 8 of these studies exclusively examined women, test of variability was significant (p > .55).4 These studies fur- of 3 examined men, and of 4 examined both men and women. ther underscore the potential importance of examining familial In 11 studies, researchers examined social support through ex- sources of social support in studies of cardiovascular regulation. perimental manipulation. Researchers of the remaining 4 stud- The studies summarized in Table 2 suggest that structured ies assessed naturalistic levels of social support. Of these 4 stud- interactions with others may also produce beneficial effects on ies, 3 studies' researchers examined functional measures of sup- cardiovascular function. However, these results may not simply port, and 1 examined a combined index of structural and be a function of the intervention discussion because such struc- functional support. tured interactions appear to generalize to others in one's net- The laboratory studies in Table 4 collectively suggest that so- work (e.g., Andersson, 1985). Therefore, the studies in Table 2 cial support may reduce cardiovascular (or autonomic nervous may produce part of their effects by increasing social compe- system) reactivity to acute psychological stress. One salient fea- (text continues on page 510) tence or the perceived importance of social interactions in one's social network (Sallis et al., 1987). At a more behavioral level of analysis, part of the association 4 We should note that 4 of the 5 studies' researchers who assessed between social support and cardiovascular function may be a familial sources of support also reported significant effectsof their stress result of health-related lifestyle factors (Umberson, 1987). For measures. Therefore, the influence of close interpersonal relationships example, social support may be associated with better cardio- and the methodological requirement suggested by S. Cohen and Wills vascular regulation because individuals high in social support ( 1985 ) are potentially confounded in these meta-analyses of buffering engage in better health practices (e.g., better diet and more effects. 506 UCHINO, CACIOPPO, AND KIECOLT-GLASER

~ ,, ~ ~ .,. ,~ ~-~ ~=_~

~,-

='t= ~. oo e.~

m k~

~'~ ~ ~

~~ e~ e~ ~~

E ~ I~ ~'~ ~..~ r~ ~.~ ~ m < m

II "~ II II g

~o ~ ~ .- .~.~ - o~ ~ SOCIAL SUPPORT AND PHYSIOLOGY 507

~ ~..~_,

¢-,,~

~'= o

"" o'o "~ ~'~ =~'~ ~ "~ ~ - ~e g s.r.

.~ ~ oNo "~"O O ~ ~ O0 .~ •~ "o o ~ ~ < < ~ m

~2 ~8.= °°

.=- .~

- ',£8. ~'

!~ .~ u~J~ _~ :~ ~N ~

Z 508 UCHINO, CACIOPPO, AND KIECOLT-GLASER

,~ 0 ~ E ~ 0 0 r~ 0 -~.~

ID. ~D ~- ~ ~

P-,,~

0 0 0 e~ o ,~ "~ ,.~ i~ ~ o'~ 0 0 ~.~ i ~N g ~ E.~ ,~ ~ o ~.- _ ~ ~ ,,:::~

8~ 0,,~0 r~ 8~

~ 0 ~o 0 0 r~ 0

~N~e e.a2

~.~ ,~ ~- ~ o

• , :~ ~I *.~ ~ .. e~ "~" ~ ~-, ~ p. I:::: .... ,~ ~-, ,'~

e m~ SOCIAL SUPPORT AND PHYSIOLOGY 509

•~ m~ o'~,~ ~ =~#,~ ~ .- I~ r,.tJ . ~- ~ o'~ o~ oO #, ,£o o~ =_ 8 ~ ~.~~'~ i °

~ 0~..~ c--~ x

~r~ .= ~ ~ ~ .~, .~ ~ ~~ C ..t.~

U:; o= ir ~,.,4 o~ 0

O~ .£o,.

Zo <=o

r~ ~g

I ~ £',,I i. ~. 0~

~.~ •~,~ ~o ~..~.~ g;

• .~ Z=m ~.=. ~ ~. ¢, 510 UCHINO, CACIOPPO, AND KIECOLT-GLASER

ture of the studies summarized in Table 4 is that strong results were women, however, the experimenter in Kamarck et al. was a emerge with laboratory paradigms that manipulate the sup- man, whereas the experimenter in Edens et al. was a woman. The portive functions of social relationships (Gerin, Pieper, Levy, gender incongruency in the Kamarck et al. study may have con- & Pickering, 1992; Kiecolt-Glaser & Greenberg, 1984; Lepore, tributed to increased evaluation apprehension and threat in the 1995; Lepore, Allen, & Evans, 1993), rather than simply relat- alone condition. Consistent with this reasoning, cardiovascular re- ing naturalistic social support measures to laboratory stressors activity levels in the alone condition for Edens et al. were lower (e.g., Boyce & Chesterman, 1990; Tardy, Thompson, & Allen, than those reported in the Kamarck et al. study. In fact, as sug- 1989). The 4 studies whose researchers manipulated the sup- gested by Edens et al., had reactivity levels in the alone condition portive function of social relationships and examined its effects been comparable with Kamarck et al., these results may have rep- on cardiovascular reactivity were characterized by a significant licated Kamarck et al?s. This interpretation is also consistent with combined test of significance (z = 4.00, p = .00003, fail-safe n the results of Snydersmith and Cacioppo (1992) who found results = 19.62) and an effect size ofr = .28. No test of variability was similar to Edens et al. using a female experimenter interacting with significant (p > .32). Such experimental studies provide strong female participants. evidence linking functional aspects of social support to changes Finally, 2 studies (Gerin et al., 1995; Kamarck et al., 1995 ) in cardiovascular parameters. For example, Gerin et al. had suggest more specific conditions under which a friend is likely participants engage in a discussion task in the presence or ab- to be associated with lower cardiovascular reactivity compared sence of social support. In the no social support condition, two with the alone condition during acute psychological stress. In confederates attacked the participant's views, while a third con- particular, in both of these studies, researchers manipulated the federate sat silently. The discussion task was identical for the degree of threat or stress present while the participants per- social support condition, however, the third confederate now de- formed the psychological tasks. Results of these studies revealed fended the participant's views. Results of the study revealed that that buffering effects of the friend condition relative to the alone social support during the conflict discussion was uniformly as- condition are likely to be found when the situation is especially sociated with lower increases in heart rate, SBP, and DBP. threatening or stressful. In general, the studies summarized in Although the laboratory studies are generally consistent with Table 4 provide evidence on the boundary conditions (i.e., de- the notion that social support may reduce ANS reactivity to gree of evaluation and threat) under which a friend may reduce acute psychological stress, studies that have operationalized so- cardiovascular reactivity to stress. cial support by having participants perform a stressful task in The laboratory studies linking social support to phasic the presence of a relationship assumed to be supportive (e.g., changes in cardiovascular function provide evidence that social friend) differ in the pattern of results obtained. In such studies, support may be important in reducing cardiovascular reactivity although the presence of a friend is associated with reduced to psyehosocial stressors. The laboratory studies provide a po- ANS reactivity compared with the presence of a stranger tentially important conceptual link to the studies summarized (Edens, Larkin, & Abel, 1992; Snydersmith & Cacioppo, in Tables 1-3 that focused on tonic blood pressure in middle- 1992), results regarding the presence of a friend compared with aged to older adult community samples. More specific, social an alone condition are inconsistent. Researchers of 7 studies support may be conceptualized as a stable individual difference examined participants' ANS reactivity to acute psychological variable (I. G. Sarason, Sarason, & Shearin, 1986), therefore, stress in such studies. Researchers of 3 studies have reported laboratory paradigms may provide a model of how social sup- that ANS reactivity to brief psychological stress was lower when port operates in everyday life. The laboratory studies suggest a friend was present than in the alone condition (Gerin, Milner, that the higher cardiovascular reactivity seen in situations in- Chawla, & Pickering, 1995; Kamarck, Annunziato, & Ama- volving low social support may translate to gradual elevations teau, 1995; Kamarck, Manuck, & Jennings, 1990). In contrast, in tonic blood pressure across the lifespan. In a relevant study, researchers of 3 studies found no differences between these con- Light, Dolan, Davis, and Sherwood (1992) examined the use of ditions (Edens et al., 1992; Sheffield & Carroll, 1994; Sny- cardiovascular reactivity to acute stressors in predicting subse- dersmith & Cacioppo, 1992 ), and K. M. Allen, Blascovich, To- quent tonic blood pressure levels. Results revealed that individ- maka, and Kelsey ( 1991 ) reported greater ANS reactivity in the uals high in heart rate reactivity evidenced elevated tonic blood friend condition compared with the alone condition. However, pressure 10 to 15 years later compared with individuals low in as suggested by Allen et al. (1991), the presence of a friend heart rate reactivity. Significantly, these groups were initially may have been associated with increased reactivity due to the characterized by comparable tonic blood pressure levels. These evaluative nature of the situation, compared with Kamarck et data are consistent with animal models of hypertension, sug- al. (1990). Consistent with this notion, Allen et al. (1991) gesting that exaggerated cardiovascular reactivity influences the found that the friend condition was associated with a faster task development of cardiovascular disorders (Folkow, Hallback, rate and more task errors compared with the alone condition, Lundgren, Sivertsson, & Weiss, 1973; Hallback & Folkow, whereas Kamarck et al. (1990) found no such performance 1974; Manuck, Kaplan, & Clarkson, 1983). differences. The differences between Kamarck et al. (1990) and Edens et al. Potential Mechanisms Linking Social Support to (1992) are more difficult to resolve. However, one explanation Reduced Cardiovascular Reactivity may lie in the interactions operating in the alone conditions. Al- though physically alone, participants in these conditions realized The laboratory studies are suggestive of several different that their performance was being visually and physiologically mechanisms linking social support to reduced cardiovascular monitored by the experimenter. All participants in both studies reactivity. At a social psychological level, nonevaluative friends SOCIAL SUPPORT AND PHYSIOLOGY 511 appear to buffer cardiovascular reactivity to highly threatening cial support to cardiovascular indices during acute psychologi- psychosocial stressors compared with the stranger or alone con- cal stress reveal that results were not a function of depression dition. One should note, however, that most of these studies (Uchino et al., 1992) or environmental stress (Knox, 1993). used undergraduate samples, and friends may be a particularly Whether this lack of mediation by psychological reactions re- important source of support in such younger populations. fleets methodological issues related to the retrospective nature These studies are also suggestive of the particular dimensions of these assessments or the possibility that other unmeasured or functions of social support that may reduce cardiovascular psychological processes may be important (e.g., feelings of con- reactivity in response to certain stressors or situations. In par- trollability and intrusive thoughts) is beyond the scope of the ticular, the dimension of emotional support appears to be a sa- existing literature. Future research is necessary to elucidate the lient feature operating in these experiments to reduce cardio- potential psychological mechanisms operating in laboratory re- vascular reactivity (Edens et al., 1992; Gerin et al., 1992, 1995; activity paradigms. Kamarck et al., 1990, 1995; Kiecolt-Glaser et al., 1985; Lepore, 1995; Lepore et al., 1993; Snydersmith & Cacioppo, 1992). Social Support and Endocrine Function These findings do not suggest that other aspects of social support (e.g., appraisal support) are unimportant because most of Endocrine literally refers to the internal secretion of biologi- the studies in Table 4 have used stressors in which emotional cally active substances. These biologically active substances, or support may be especially important. Appraisal support, for in- hormones, are usually defined as substances that are released stance, cannot be effective if there is no opportunity for discus- from an endocrine gland (e.g., pituitary) into the blood stream sion in these studies. Therefore, consistent with past research to act on a distant target tissue site (see Greenspan & Baxter, (S. Cohen & Wills, 1985; Cutrona & Russell, 1990) and the 1994). Hormones can act in either an autocrine (i.e., act on the studies summarized thus far, the laboratory paradigms suggest same cells that produced them) or paracrine (i.e., act on cells that particular functions of social support are more effective other than those that produced them) fashion. The major endo- when there is a reasonable match with the demands of particu- crine glands include the thyroid, pituitary, pancreas, adrenal lar situations. On the basis oftbe studies presented in Tables 1- medulla, adrenal cortex, testes, and ovaries. 3, the match between the support dimension and stressor may The most commonly examined endocrine measures in stress also need to be considered in the context of cultural and gender- research include the catecholamines (e.g., norepinephrine and related processes. epinephrine) and cortisol. Epinephrine (EPI) is produced and In many of these studies, data were provided regarding po- released into the blood stream from the adrenal medulla. Nor- tential behavioral mechanisms linking manipulated or natural- epinephrine (NE) is also synthesized in the adrenal medulla istic social support with lowered cardiovascular reactivity. The but is largely produced in the central nervous system and pe- studies in Table 4 suggest that behavioral factors, such as poten- ripheral sympathetic nerves. Catecholamines bind to adrener- tial health-related variables (e.g., Uchino, Kiecolt-Glaser, & gic receptors where their effects are to increase oxygen and heat Cacioppo, 1992) and performance indices during the tasks consumption and activate glucose and fat from storage areas in (Edens et al., 1992; Gerin et al., 1992, 1995; Kamarck et al., the body. Some important physiologic changes in response to 1990, 1995; Lepore, 1995; Lepore et al., 1993; Snydersmith & catecholamines include increased heart rate, increased myocar- Cacioppo, 1992; Uchino et al., 1992) were not responsible for dial contractility, changes in vascular resistance, and regulation the results obtained in these studies. Therefore, consistent with of renin secretion from the kidneys. the correlational studies reviewed earlier, the laboratory studies Cortisol is a cholesterol-derived steroid, secreted from the suggest that behavioral factors are not necessary for associations adrenal cortex where it acts through the cytosolic glucocorti- between social support and cardiovascular reactivity. coid receptor protein. Cortisol secretion is a result of complex Several studies in Table 4 examined the possibility that spe- regulatory processes and involves corticotropin releasing hor- cific psychological and emotional reactions during the labora- mone (CRH) and ACTH. The biologic process is initiated, in tory tasks mediated the effects of social support on cardiovas- part, by the release of CRH from the hypothalamus that signals cular reactivity. These studies' researchers generally reported the pituitary gland to secrete ACTH. ACTH then signals the that participants' emotional reactions during the experiment adrenal cortex to release cortisol and later serves to regulate the did not differ as a function of the social support conditions amount ofcortisol secreted. Once released, cortisol has a vari- (Edens et al., 1992; Gerin et al., 1992; Kamarck et al., 1990, ety of metabolic effects, including increased glucose metabo- 1995; Lepore et al., 1993; Snydersmith & Cacioppo, 1992; lism and down regulation of immune function. Uchino et al., 1992) and therefore do not appear to be respon- An examination of endocrine function is important because sible for the pattern of results in these studies. In an illustrative of its association with both the cardiovascular and immune sys- study, Lepore et al. (1993) had participants perform a speech, tems. As reviewed earlier, the catecholamines play an important while in one condition a confederate made comments designed role in cardiovascular regulation. For instance, catecholamines to reflect emotional support. In contrast, the confederate in the bind to alpha-adrenergic receptors and lead to constriction of low emotional support condition was reserved and inattentive. arterial smooth muscles. Furthermore, catecholamines appear Results indicated that the social support speech condition was to be toxic to endothelial cells of the arteries (Krantz & Ma- associated with lower SBP and DBP reactivity than the no sup- nuck, 1984). Damage to the endothelial cells of the arteries is port speech condition. In addition, results revealed that these thought to be an important initiating factor in the development conditions did not differ in terms of perceived stress during the of coronary atherosclerosis (Gorlin, 1976). speech. Data from 2 additional studies relating naturalistic so- Clearly, important interactions exist between the endocrine and 512 UCHINO, CACIOPPO, AND KIECOLT-GLASER immune systems (Ader, Felton, & Cohen, 1991 ). Lymphocytes chronic or long-term alterations in endocrine function but are appear to have beta-adrenergic, glucocorticoid, and opioid recep- relatively insensitive to short-term bursts of catecholamines. tors (Plaut, 1987; Sibinga & Goldstein, 1988 ). More significantly, Finally, the method of obtaining samples for endocrine as- hormones influence numerous aspects of immune function. For sessments may contribute significant sources of variance. If instance, the acute release of catecholamines produce transient plasma samples are assessed, blood draws that are obtained increases in natural killer (NK) cell lysis (i.e., an in vitro model of from a single needle stick can elevate catecholamines and make the functional ability of NK cells to lyse or destroy tumor or virus- relationships with social support more difficult to detect. There- infected ceils; Tonnesen, Christensen, & Brinklov, 1987) and de- fore, the use of a catheter with a rest period that allows for creases in the proliferative response to mitogens (i.e., an in vitro changes due to venipuncture to subside is typically recom- model of the functional ability of lymphocytes to proliferate when mended (Baum & Grunberg, 1995). The use of urinary sam- faced with a challenge; Crary et al., 1983). In addition, glucocor- pies is associated with different methodological considerations. ticoids have regulatory effects on immune responses, including de- For instance, compliance with 24-hr urine samples may be creased NK cell lysis (Gatti et al., 1987 ) and mitogen proliferation problematic so that 15-hr overnight samples are useful (Baum (Rupprecht et at., 1990). Prolactin, growth hormone, estradial, & Grunberg, 1995). testosterone, opioids, CRH, and ACTH also have immunomodu- As evidenced by inspection of Table 5, the association be- latory effects (see reviews by Blalock, 1989; Carr, 1991; and Cupps tween social support and endocrine function has not been well & Fauci, 1982). studied. Only 10 studies included endocrine assessments: 4 ex- The measurement reliability of commonly examined endocrine amined catecholamines, 4 examined cortisol, and 2 examined measures (e.g., catecholamines and cortisol) is well documented. both. In 4 studies, the investigators assessed hormones in urine, For instance, one of the most accurate and cost-effective tech- in 1 study hormones were assessed in saliva, and in 1 study hor- niques, radioimmunoassay (RIA), uses specific binding sub- mones were assessed in breast milk. In addition, 4 studies as- stances (e.g., antibody) with radioactive tracers to quantify endo- sessed plasma endocrine measures, and only 1 study (Knox et geneous hormones in bodily fluids (see Bissette & Ritchie, 1992; al., 1985 ) mentioned that they had drawn blood from a catheter Chard, 1982 and Jaffe & Behrman, 1979). RIA has very good that was inserted before a rest period. As noted earlier, this may specificity and sensitivity and is characterized by low intra-assay be important because the needle stick associated with a single variability (i.e., typically less than 1% to 5%). However, the in- blood draw can elevate catecholamines and make relationships terassay variability may vary depending on factors such as the sta- more difficult to detect. bility of solutions and radioactive tracers, incubation temperature In regard to sample characteristics, 4 studies examined only and duration, and changes in the sample matrix and technicians men, 2 studies examined only women, and 4 studies used both (Bissette & Ritchie, 1992). Nevertheless, quality control proce- men and women. Social support was also operationalized in sev- dures (e.g., monitoring standard curve parameters) are typically eral ways: 6 studies assessed naturalistic social support levels, implemented such that interassay variability below 10% may be and 4 studies manipulated social support (i.e., interventions and laboratory reactivity studies). Of the 6 studies that assessed achieved (Bissette & Ritchie, 1992). naturalistic social support, 4 assessed functional aspects of sup- Little data exist on the temporal stability of interindividual port, 1 assessed both structural and functional aspects of sup- variations in endocrine function. 5 There are a variety of meth- port, and 1 assessed a combined index of both structural and odological issues worth considering in examining the temporal functional aspects of support. stability of endocrine measures (see reviews by Baum & Because of the small number of studies and differences in the Grunberg, 1995; and McKinnon, Baum, & Morokoff, 1988). operationalization of social support (e.g., naturalistic social First, there are diurnal variations associated with both catechol- support and laboratory manipulations), we performed a quali- amines and cortisol (Akerstedt & Levi, 1978; Greenspan & tative analyses of the studies whose researchers examined social Baxter, 1994). Therefore, if only a limited number of samples support and endocrine function. To this point, researchers of 5 are drawn, it is important to collect such samples at approxi- of the 6 studies reported an association between social support mately the same time to reduce this extraneous and potential and catecholamines levels (Arnetz, Edgren, Levi, & Otto, 1985, confounding source of variance. In addition, even if samples are Ely & Mostardi, 1986; Heming, Baum, Gisriel, & Gatchel, obtained at the same time, the point in the diurnal cycle that 1982; Knox et al., 1985; Seeman, Berkman, Blazer, & Rowe, endocrine function is assessed may be important. For instance, 1994). In an early study, Fleming et al. examined the influence cortisol has a diurnal cycle that peaks during the morning hours of perceptions of support in participants exposed to the chronic (Greenspan & Baxter, 1994) so that relationships with social stress of Three-Mile Island and in control participants. Results support may be more difficult to detect at this time. Thus, one revealed that individuals low on social support were uniformly may need to examine the relationship between social support characterized by higher urinary NE levels than individuals high and cortisol across the full range of the diurnal cycle. on social support (a similar but nonsignificant trend was found Second, it is important to distinguish whether endocrine for EPI). measures are assessed in plasma or urine (Baum & Grunberg, Of the 6 studies whose researchers have examined cortisol 1995). This distinction is especially relevant for the assessment of plasma catecholamines that have a relatively short half-life 5 A literature search on PsycLIT (1974-1994) and Medline ( 1983- of 1-3 rain. As a result, plasma assessments of catecholamines 1994) crossing the keywords endocrine or neuroendocrine with psycho- provide data on transient changes in endocrine function. Uri- metrics or reliability revealed no researchers had examined the tempo- nary assessments, however, provide information on more ral stability of the endocrine assessments discussed in this review. SOCIAL SUPPORT AND PHYSIOLOGY 513 levels, researchers of 4 have found no association with social amine levels. Researchers of 2 studies conducted under the support (Arnetz et al., 1983, 1985, 1987; Groer, Humenick, & stress-buffering model found no evidence for a buffering effect Hill, 1994). Two of these studies are interventions (i.e., Arnetz on catecholamine levels for participants faced with job termi- et al., 1983, 1987), and 1 of these studies' data suggest the in- nation (Cobb, 1974) or chronically stressed Three-Mile Island tervention successfully increased social contact (Arnetz et al., residents (Fleming et al., 1982). Moreover, both studies' re- 1983). Although the number of studies is small, these data sug- searchers found that the stressor was associated with significant gest that simply increasing social contact may be insufficient to elevations in catecholamine levels. Because of the small number affect cortisol levels, at least at nomethetic levels of analysis. of studies, more data are needed to adequately test the stress- However, there are important issues that may have prevented an buffering model on catecholamine levels. adequate test of the relationship between social support and In a small number of studies summarized in Table 5, research- tonic cortisol levels. First, idiographic analyses of cortisol ers have examined psychological processes and its relationships to changes to brief psychological stressors suggest that such social support (Arnetz et al., 1985; Fleming et al., 1982; Kirsch- changes may be heightened in individuals characterized by high baum et al., 1995). Fleming et al. found that high social support cardiac sympathetic reactivity (Cacioppo et al., 1995). An ex- was associated with lower depression and anxiety, along with lower amination of individual differences, therefore, may provide ad- feelings of alienation. However, analyses specifically examining ditional information about the effects of social support on cor- mediating or moderating mechanisms were not conducted be- tisol levels. cause they were not a primary aim of the study. In this regard, laboratory paradigms of social support may be In the study by Arnetz et al. (1985), social support was con- helpful to examine phasic cortisol response. In a recent study, ceptualized by examining neurotic boys' endocrine responses Kirschbaum, Klauer, Filipp, and Hellhammer (1995)exam- to an exciting and partly violent film in either a different school ined the influence of social support from a stranger or partner (i.e., low support) or same school (i.e., high support) setting. (i.e., boyfriend or girlfriend) on cortisol reactivity during acute Arnetz et al. found that the low support condition was associ- psychological stress. Results revealed that partner-supported ated with greater catecholamine changes in response to the film men showed significantly lower cortisol levels than the stranger- than the high support group. In addition, the low support group supported or no-support conditions. In contrast, women later preferred to review film scenes representing feeling of se- showed a trend toward greater cortisol response during the part- curity, whereas the high support group preferred to review film ner-supported compared with the other two conditions, even scenes representing danger, suggesting that increased feelings of though women reported feeling supported by their partners. anxiety may be one mechanism operating in this study. These results, however, should be considered preliminary and The laboratory social support study by Kirschbaum et al. require replication. (1995) examined perceptions of support and stress and found There is another potential reason why the intervention stud- that these psychological mechanisms could not explain the ob- ies summarized in Table 4 did not find a relationship between tained pattern of results. These results are consistent with those social support and cortisol levels. The studies summarized ear- summarized in Table 4. All in all, more research is needed on lier in Tables 1-3 on social support and cardiovascular function the relationship between social support and endocrine function, suggest that familial sources of social support may be impor- along with an examination of psychological and behavioral tant. None of the intervention studies was designed to facilitate (e.g., health-related) mechanisms responsible for these associa- familial sources of support. In addition, none of these studies' tions. Greater attention to methodological issues in examining researchers examined the potential generalization of their inter- the relationship between social support and endocrine function vention program to such support sources. (e.g., sufficient rest period after needle stick) may also prove Three of the 4 studies that did not find a relationship between useful. social support and cortisol obtained specimens primarily during the morning hours. One potential problem with a morning Social Support and Immune Function assessment is that cortisol has a diurnal cycle that peaks during the morning hours (Greenspan & Baxter, 1994). Therefore, the The immune system is the body's defense against infectious higher cortisol levels seen during morning hours may make it and malignant disease (see Borysenko, 1987; Calabrese, Kling, more difficult to detect relationships between social support & Gold, 1987; and Kiecolt-Glaser & Glaser, 1988b, 1995, for and tonic cortisol levels. It is therefore promising that Seeman basic reviews). The major organs of the immune system are the et al. (1994) examined 12-hr urinary cortisol samples and re- thymus, bone marrow, lymph nodes, spleen, tonsils, appendix, ported relationships between social support and cortisol levels and Peyer's patches. in men. The study by Seeman et al. is also one of the most meth- There is no single measure of immune function, and re- odologically rigorous because the researchers also statistically searchers in the field of psychoneuroimmunology (PNI) have controlled for a variety of potential confounding variables (e.g., used various measures to index aspects of immune function. A age, medication use, and body mass). distinction has typically been made between quantitative and functional measures of the immune system (Kiecolt-Glaser & Potential Mechanisms Linking Social Support to Glaser, 1988a). Quantitative measures can include absolute counts or percentages of certain immune cells such as helper T Endocrine Function cells, suppressor/cytotoxic T cells, and NK cells. Quantitative On the basis of Table 5, there appears to be preliminary evi- measures are typically examined because both the number and dence for a relationship between social support and catechol- (text continues on page 519) 514 UCHINO, CACIOPPO, AND KIECOLT-GLASER

~= = '~ ~ 1:: ~. "~ 'o ~ ~ ~ ",~ ~, ¢o ~o~ z o 87 ~

~z & # ~'J"~£"x---~'~''~6~'o'aa~ o~='=-"-8'-=-'~=: ..-.~o.--~,= ~o,~,,.=~.,~, "0 o

o o o o Z z z z

Q,, ~J

Z z Z--,~

"t2 "~

o o o z z z

•= ~ -4 o,~.,

r.~

~.E o xo -~

-4

,'4 C~,O E 'c.~- < < SOCIAL SUPPORT AND PHYSIOLOGY 515

1::

~ z-; r,,.- " ~,.__.~ .---2_~ ~.~ ~ r, o ~ ~' o ~ ~-. = 1::.- ~ ~,~ ." "~

_~,,o e~ : ~.., .,,.1, .... ~ 8 8.9--'U,=~ ° i~'~_o. ~ ~± Z ~ ~.~ E

.- ~, ~,1:'~ ~ o,",~o~.~ "~'~ ~ z < Z =

E "S~ 88 "= ~ =~-~. ~ , < ~- ~.~- ~>:=

0 "r' "l~ ~ '-~ ¢"*' 0 0 z Z " Z Z Z

,,.,¢

b i .~.~E'~ ,,, ~ - ~ z

= Z Z Z Z

._~; ~,~ ~ =~

~ '~, :~

o °o

~_o °

~- = ~.~.~. ~ E ~'- "~'~ 516 UCHINO, CACIOPPO, AND KIECOLT-GLASER

u., ,~

~:=2 g, z "= =-K ~.~ ~+ -~E~. 0oo ~ -o _ ~1~ o ,.. o o ='~ o ~"

~_~-~ < < < r~

~ +~ r~-t-

< E 2 ~ ~. z z g-~ z

0 0 0 o 0 0 O CJ Z z z z Z Z Z Z

0 0 0 0 0 o o z Z z z Z Z ;~ Z

"• ~'~ 1:: E ° o

Ei ~m ~ ~.~ ~ ~.~=~ ,~ ~.- ~ ~ ~=~ <

~ ~,,

~.- =.~

~'0

0 ~ 0 0 8..o ~-a ._o ~ o=

o~ 2~ SOCIAL SUPPORT AND PHYSIOLOGY 517

-~',., ~ ~ ~o ~= .= ~.~ -~ .~.;

o =o, ~ ~,7= ,','= = ~,- E,E

=,,.to o ~ o o~ ='--'~-"'~= -~,

•~ .£ ~ Rs ~ .'-"Z e# 0 ,~1 =.=,-,= . 8= .... £~=.- ,.1= c,~ t# .~ (J Z ~ [- I--

"~ ~ o < E .4~ 8 &,.-, ~ o ~ ~+ =+~+ E~,~ "~-~ ';:1 E o ~- ~ ~z~ ,5

0 0 0 0 0 Z Z Z Z Z Z

Z Z Z Z Z

e, o

. = > ~ = ~ ~ o=

.~- ~

~ ~ ~ ,,< ,,

~- ~= ,~ la, e-4 .~ ~

r.~ ~ ~ ~m

..qo

L3 [..3 18 UCHINO, CACIOPPO, AND KIECOLT-GLASER

0 m

.~ .~,.~ ± ~~,,.~'~ ~~ ~ ~ ~-~ i.~ ~ pj~ ~.~. ~ ~~='= ;_~ "- ~m ~-~.~I m~...~ ~~ ~ ,..~ ~ ~* ~'~ ~ ~. ~ a.~. ~' ~*0 ~.~-~'~ ~ ~-~ ~.~~ ~..~ ~. .. ~ ~,-."~ ~. ~.-~o~ ~-~, "~ "- ~d .~,,~ .oR~.~ 'e ~.~+~.=~ '~ ~. z ~

~~; .~

z z

~..~_ ~'~ ~,-.~ ~ 0~ =-~='~. ~.e.~ ~ = ~ ,,~ ~- ~, e~ = :'~ ~ .~ _=__, ~.~ ,:o ,~ ~t.~ ~ ~ ~ ~ ~,~ .-~.~-~=~

~.~; .~'~..+ - r~.~ 0 I~

~ ~ ~ ~ ,-" ",,~ 0 ~.~'~

~" Z ~.=~

._~ ,,-~- ~ II ~ o1~'rgl~

8 e~ ~.~

'~ "~",i' ~ .~ . ,~ ~.- .~ k~ i k. SOCIAL SUPPORT AND PHYSIOLOGY 519 relative balance of immune cells (e.g., ratio of helper T cells aspects of immune function would be in a stronger position to to suppressor/cytotoxic T cells) are important in mounting an make such conclusions by examining the pattern of immune effective immune response (Herbert & Cohen, 1993b). Func- changes across assays. tional measures examine the performance of certain immune We found 19 studies whose researchers examined the associ- cells. One common measure is the blastogenic response oflym- ation between social support and aspects of immune function phocytes to the plant mitogens concanavalin A (Con A) and (see Table 5 ).7 Eight studies used only functional assays, 6 stud- phytohemmaglutinnin(PHA). Blastogenesis provides an in vi- ies only quantitative assays, and 5 studies both quantitative and tro model of lymphocyte proliferation in response to antigens. functional immunological measures. Nine studies were per- In general, greater proliferation is interpreted as a better im- formed with middle-aged or older adult samples (Mage > 40). mune response. In addition, measures of NK cell lysis are taken Nine studies tested both men and women, 6 only women, and 4 by incubating NK cells with a tumor cell line and examining only men. In addition, l0 studies were correlational, 7 prospec- the ability of NK cells to lyse (i.e., destroy) tumor cells. NK tive, and 2 experimental interventions. Of the 17 studies whose cells appear to play an important role in the body's defense researchers assessed naturalistic social support, only 4 included, against virally infected and malignant cells (Whiteside & Herb- in part, structural measures of support. erman, 1994). The results of the 19 studies summarized in Table 5 are con- The measurement reliability of immunological data varies sistent with the notion that higher social support is associated across assays and laboratories. Issues such as the day on which with better immune system function (e.g., greater NK cell an individual's immunological data are analyzed can contribute lysis). To summarize, 12 studies' researchers reported evidence significant variance for some assays. Considerable work has been that social support was associated with aspects of immune func- done to ensure the measurement reliability of some assays across tion, and 7 reported no relationship. A meta-analysis of 9 stud- laboratories and times (e.g., percentages of T lymphocytes and ies whose researchers examined the association between social subpopulations; Gelman, Cheng, Kidd, Waxdal, & Kagan, 1993; support and functional immune measures revealed a reliable Paxton et al., 1989; Schenker et al., 1993). Similarly, studies per- combined test of significance (z = 4.38, p = .000006, fail-safe n formed after implementation of quality control procedures to es- = 54.90). The effect size associated with these studies was r = tablish reliability and reduce daily variability of NK cell lysis show .21, and no test of variability was significant (p > .49).s that NK activity is also characterized by adequate measurement reliability (Whiteside & Herberman, 1994). However, researchers 6 A literature search on PsycLIT(1974-1994) and Medline ( 1983- in only a few studies have examined the temporal reliability of 1994) crossing the keywords immune or immunology with psychomet- common measures of immune function. Researchers reported rics or reliability revealed no additional researchers who had examined data on the 3-week temporal stability of several quantitative and the temporal stability of the immune assessments discussed in this functional immune measures in response to a stressful speech task review. (Marsland, Manuck, Fazzari, Stewart, & Rabin, 1995 ). Marsland 7 Our initial search produced 24 studies whose researchers examined the relationship between social support and immune function. We ex- et al. found that most baseline quantitative measures of immunity cluded analyses of 6 studies in part or in whole for several reasons. Data (e.g., CD4+ and CD8+ cells) evidenced significant temporal sta- on the relationship between social support and white blood cell or total bility (.22 -< r ~ .75). Moreover, immune responses to acute stress lymphocyte counts were not included in the present review (Arnetz et were also associated with significant temporal stability for most al., 1983; R. S. Baron, Cutrona, Hicklin, Russell, & Lubaroff, 1990; quantitative measures (.25 -< r -< .53), as well as changes in the Mclntosh, Kaplan, Kubena, & Landmann, 1993; Thomas, Goodwin, proliferative response to PHA (r = .50, p < .005 ). & Goodwin, 1985 ) because of difficulties in interpreting the signifi- According to psychometric theory, several assessments of an cance of these measures. In addition, we excluded 1 study that exam- individual across times and situations are likely to provide a ined salivary IgA (Jemmott & Magloire, 1988 ) because of methodolog- more accurate individual difference assessment of immune ical issues regarding the reliability of the salivary lgA assessment that function. Consistent with this possibility, Fletcher, Klimas, was used in this study (Herbert & Cohen, 1993a; Stone, Cox, Morgan, and Gjerset (in press) found that the proliferative re- Valdimarsdottir, & Neale, 1987 ). One study was also excluded from the review because the small number of participants (3) in their low stress sponse to PHA, pokeweed mitogen, and NK cell lysis were char- and no social support cell precluded definitive analyses (Herrera, Alva- acterized by adequate generalizability (G) coefficients (Gs > rado, & Martinez, 1988 ). .70), which increased when assessments were aggregated across s The meta-analysis consisted of 9 studies whose researchers directly times (Gs > .85 ). Unfortunately, the high costs of immunolog- examined the association between social support and functional mea- ical assays make repeated determinations impossible in many sures of immune function. To reduce problems associated with nonin- circumstances. Although the data obtained by Marsland et al. dependence, the results of Snyder, Roghmann, and Sigal ( 1990, 1993) ( 1995 ) and Fletcher et al. suggest that measures of immune re- were averaged to produce one test statistic because data were reported sponse are characterized by good temporal reliability, addi- from the same sample. Six studies were excluded from the meta-analysis tional data are needed to examine an individual difference as- because it was determined a priori that intervention studies and popu- sessment of immune function.6 lations with individuals having HIV were associated with inconsistent effects. These studies are discussed in detail later in the text. There are additional issues involved in the interpretation of In the meta-analysis, we examined functional immune measures be- measures of immune function that may be useful to discuss. As cause there were only 4 remaining studies whose researchers examined mentioned earlier, there is no single generally accepted measure the association between social support and quantitative measures of im- of immune function. As a result, the interpretation of a single mune function. The interested reader is referred to Table 5 for a sum- measure of immune function as representing a down regulation mary of the relationship between social support and quantitative im- may be difficult to determine. Studies that measure multiple mune measures. In addition, there were only 2 remaining studies that 520 UCHINO, CACIOPPO, AND KIECOLT-GLASER

Two of the studies that did not find an association between (Effros & Walford, 1987). Excluding the intervention studies social support and immunological data were interventions de- discussed earlier, it is important to note that the association be- signed to facilitate social interactions (Arnetz et al., 1987; Kie- tween social support and functional measures of immunity is colt-Glaser et al., 1985). The Kiecolt-Glaser et al. ( 1985 ) inter- consistent in older adults. A meta-analysis of 7 studies in mid- vention documented increased social contact and interactions. dle-aged to older adult populations confirmed this hypothesis (z However, keep in mind that the sample sizes in 2 of these studies = 4.27, p = .000009, fail-safe n = 40.30), with an effect size of are relatively small (n < 45.00). In addition, we reviewed evi- r = .23. No test of variability was significant (p > .35). dence earlier suggesting that familial sources of support may Although 9 studies' researchers examined both men and be important. It is possible that interventions aimed directly at women, 8 studies' researchers did not report analyses aimed at increasing familial contact and support may yield even larger examining potential gender differences. In the only study with effects sizes on immune function. data on potential gender differences, Thomas, Goodwin, and Two of the studies that did not find a significant association Goodwin (1985) found that the availability of a confidant was between social support and aspects of immune function exam- associated with a stronger proliferative response to PHA for ined men who were HIV+ (Goodkin et al., 1992; Perry, Fish- women but not men. However, the correlations were in the same man, Jacobsberg, & Frances, 1992). There are important meth- direction, and no statistical test was performed to directly test odological reasons that might explain a lack of an association the difference between men and women. In addition, Thomas et between social support and immune function in individuals al. provided a conservative test of the effects of social support HIV+. For instance, stage of disease, age, gender, drug abuse, on immune function, as they statistically controlled for psycho- and health behaviors are potentially important confounding logical distress as well as potential health-related variables (e.g., variables ( Ironson et al., 1994). alcohol consumption). However, 2 recent studies have reported an association be- Similar conceptual issues exist in the research examining so- tween social support and CD4+ counts (a marker of HIV cial support and immune function as in the research reviewed progression) in men with HIV (Persson, Gullberg, Hanson, earlier. Only 2 of the studies summarized in Table 5 conceptu- Moestrup, & Ostergren, 1994; Theorell et al., 1995). In 1 pro- alized social support as a multidimensional construct and re- spective study with data across a 5-year period, Theoreli et al. ported analyses regarding a relatively specific dimension of so- found that the availability of social and emotional support pre- cial support (R. S. Baron et al., 1990; Persson et al., 1994). dicted subsequent changes in CD4+ counts in a representative Although 3 additional studies used multidimensional social Swedish sample of men with HIV. Results revealed that high support measures, results were only reported on the total scale and low social support groups did not differ in CD4+ counts (e.g., Glaser, Kiecolt-Glaser, Bonneau, Malarkey, & Hughes, during the early years of the study. However, the prediction of 1992; Goodkin et al., 1992; Perry et al., 1992). An examination CD4+ counts as a function of social support was evident during of relatively distinct dimensions of social support may have Years 4 and 5 of the study. For instance, during Year 5 of the revealed greater specificity (Glaser et al., 1992) and stronger study, individuals high in social support showed a -37% change associations between social support and immune function in CD4+ counts, whereas individuals low in social support (Goodkin et al., 1992; Perry et al., 1992) due to a better match showed a -64% change in CD4+ counts (Theorell et al., 1995 ). between the sample needs and the support resource. Note that Perry et al. (1992) reported null findings on the relationship between social support and CD4+ counts only up to Potential Mechanisms Linking Social Support to Year l of their study. These preliminary prospective data sug- Immune Function gest that social support may influence the progression of HIV infection and provide evidence on the utility of such long-term The studies summarized in Table 5 suggest that social sup- prospective designs. An important implication of these data is port is associated with better immune function. These results that if a researcher was to only examine the relationship be- and a recent meta-analysis conducted by Herbert and Cohen tween social support and CD4+ counts later in the stage of dis- (1993a) provide converging evidence for the effects of social ease, information on the longer length of time that individuals support on physiological function. However, Herbert and Co- high in social support took to get to that stage would be lost.9 hen only examined aspects of immune function and focused on A population of particular interest in this review is older social stressors involving the loss or disruption of interpersonal adults because social support may be especially important for resources (e.g., bereavement and marital conflict). these individuals (House et al., 1988). Alterations in immune As in the review of social support and cardiovascular function function may have significant consequences in this population (see Tables 1-3), the studies summarized in Table 5 suggest as aging is associated with a down regulation of immune func- that close relationships, such as familial ties, may be a particu- tion (Goidl, 1987; Goodwin, Searles, & Tung, 1982; Roberts- larly important source of social support. Researchers of 2 stud- Thomson, Whittingham, Youngchaiyud, & Mackay, 1974; ies assessed social support specific to close relationships (Levy Schleifer, Keller, Bond, Cohen, & Stein, 1989), and infectious illnesses are the fourth leading cause of death in the elderly 9 We should note that a typical strategy in such designs might be to statistically control for the length of time since illness. However, if the interaction between this covariate and social support were significant, it included a structural assessment of support. Therefore, focused com- would invalidate the use of this statistical control procedure (J. Cohen parisons between structural and functional measures of support were & Cohen, 1983) but accurately reflect the findings of Theorell et al. not performed. (1995). SOCIAL SUPPORT AND PHYSIOLOGY 521 et al., 1990; Thomas et al., 1985), and both found social sup- not appear to be necessary for an association between social port to be related to aspects of immune function, including a support and immune function. However, these findings should stronger proliferative response to PHA and greater NK cell lysis. be taken as preliminary, given the restricted number of health- Researchers of 4 of the studies in Table 5 directly tested the related practices assessed and the lack of reported data on the buffering model of social support (Goodkin et al., 1992; Kie- validity and reliability of such assessments in many of these colt-Glaser, Dura, Speicher, Trask, & Glaser, 1991; Snyder, studies. Roghmann, & Sigal, 1990, 1993). Of these studies, only Kie- Psychological factors such as levels of stress and depression colt-Glaser et al. ( 1991 ) reported a reliable effect of stress on have reliable effects on immune function (Herbert & Cohen, immune function. As noted earlier, S. Cohen and Wills (1985) 1993a, 1993b). Therefore, part of the association between so- argued that this is a methodological requirement for an ade- cial support and immune function may be mediated by these quate test of the buffering model. More important, Kiecolt-Gla- factors. Researchers of 3 studies in Table 5 reported data relat- ser et al. found evidence for a buffering effect of social support ing to potential psychological mechanisms responsible for the on immune function. However, because of the small number of associations between social support and immune function studies, more data are needed to adequately test the buffering (R. S. Baron et al., 1990; Glaser et al., 1992; Kiecolt-Glaser et model on immune function. al., 1991 ). Baron et al. (1990) found that the associations be- Two of the studies summarized in Table 5 conceptualized so- tween social support and immune function were not mediated cial support as a multidimensional construct and reported anal- by life events. In addition, Baron et al. (1990) and Kiecolt-Gla- yses regarding relatively specific dimension of social support ser et al. ( 1991 ) found that depression was not mediating the (R. S. Baron et al., 1990; Levy et al., 1990). Levy et al. exam- associations between social support and immunity. Finally, Gla- ined the dimension of emotional support from a spouse (or in- ser et al. (1992) reported evidence indicating that anxiety levels timate other) and emotional support from one's doctor and were not responsible for the associations between social support found both to be associated with greater NK cell lysis in cancer and immune function. Therefore, although health-related be- patients. In a study of spouses of cancer patients, Baron et al. haviors, depression, and life stress have reliable effects on as- used the social provisions scale and found that higher levels on pects of immune function, these factors do not appear to be all support dimensions (i.e., guidance, reliable alliances, reas- major pathways explaining the associations between social sup- surances of worth, social integration, attachment, and opportu- port and immune function. nity for nurturance) were equally and significantly associated with a stronger proliferative response to PHA and greater NK Discussion cell lysis. As noted by Baron et al., caring for a spouse with Social support has been linked to lower rates of morbidity cancer may result in a mobilization of one's support network, and mortality from diverse disease processes and endpoints. such that there was little differentiation among support compo- Therefore, the major aims of this review were to examine the nents. Consistent with this possibility, Baron et al. reported high evidence linking social support to multiple aspects of physio- intercorrelations among the components of support. logical function and to characterize the potential mechanisms Levy et al. (1990) suggest that emotional support may be one responsible for these covariations. To the best of our knowledge, dimension of social support that is associated with immune this is the first comprehensive review on this topic. The present function. Researchers of 4 additional studies also assessed, in review indicates that there is relatively strong evidence linking part, emotional support (also see R. S. Baron et al., 1990). Kie- social support to aspects of the cardiovascular, endocrine, and colt-Glaser et al. ( 1991 ) and Esterling, Kiecolt-Glaser, Bodnar, immune systems. These data are consistent with research sug- and Glaser (1994) used a composite index of emotional and gesting that the formation and disruption of social relationships tangible support. Snyder et al. ( 1990, 1993) used a composite have important immunological and endocrinological sequalae index of emotional and informational support. A meta-analysis in nonhuman primates and humans (Coe, 1993; Gunnar, 1992; of these studies revealed a significant combined test of signifi- Herbert & Cohen, 1993a). More important, the physiological cance (z = 4.02,p = .00003, fail-safe n = 24.90). The effect size systems reviewed may play important roles in the leading causes associated with this test was r = .26, and no test of variability of death in the United States, including cardiovascular disor- was significant (p > .44). These data suggest that emotional ders, cancer, and respiratory illnesses. 10 Conceptual and meth- support may be at least one important aspect of social support in predicting immune function. Additional research is needed, ~0 C. E. Smith, Fernengel, Holcroft, Gerald, and Marien (1994) con- however, that directly compares the predictive utility of specific ducted a meta-analysis on the effects of social support on various health dimensions of social support. measures, including physical and stress-related outcomes. They opera- At a behavioral level of analysis, part of the association be- tionalized stress outcomes as reports of negative life events, conflict or tween social support and immune function may be due to their distress, and laboratory measures, such as catecholamine levels. Physi- effects on potential health-related variables (Kiecolt-Glaser & cal health status was operationalized as subjective states, such as symptoms and signs, and as objective data, such as weight loss, activities of Glaser, 1988b). Several researchers assessed the effects of po- daily living, blood pressure, blood glucose, and reports of sexual activity tential health-related behaviors and found that the associations posthysterectomy. The results of the meta-analyses revealed effect size between social support and immune function were significant estimates ranging from .01 to .22. Smith et al. concluded that the rela- even when statistically controlling for health practices (Thomas tively small effect sizes suggest that the relationship between social sup- et al., 1985; also see Theorell, Orth-Gomer, & Eneroth, 1990). port and health may not be significant or generalizable. These data are consistent with results reviewed earlier on social There are several issues that warrant discussion regarding the C. E. support and blood pressure, suggesting that such behaviors do Smith et al. (1994) meta-analysis. First, Smith et al. did not present any 522 UCHINO, CACIOPPO, AND KIECOLT-GLASER odological issues were also raised regarding the associations be- support was difficult to characterize in the present review, but tween social support and physiological processes. We now turn consistent with Seeman et at. there was evidence that emotional to these issues. support has reliable effects on physiological function. Given the One basic and recurring issue in the social support literature conceptual advantages to a multidimensional approach, future relates to the measurement of social support (Barrera, 1986; S. researchers might profitably use such an approach and report Cohen & Wills, 1985; Heitzmann & Kaplan, 1988; Orth- data on both specific dimensions as well as overall levels of so- Gomer & Unden, 1987; Tardy, 1985; Winemiller et al., 1993). cial support. Questions remain about the factor structure of social support An important issue examined in the present review relates to and the temporal stability and psychometric properties of such the potential psychological mechanisms linking social support assessments. In the present review, few studies' researchers re- to physiological function. A large literature has documented ported data pertaining to the psychometric properties of their psychological antecedents and consequences of social support social support measures. Given the heterogeneity in the mea- (e.g., B. R. Sarason, Sarason, & Pierce, 1989). However, the sures of social support covered in this review, psychometric data present review indicates that most studies' researchers have not may help clarify reliable relationships. reported data on such potential mechanisms, and when such An additional measurement issue concerns the specific data were presented, appropriate statistical procedures for ex- sources of social support. The present review indicates that fa- amining mediation and moderation were typically not per- milial ties appear to be an important source of social support formed (see R. M. Baron & Kenny, 1986 ). One interesting pre- to consider in studies of physiological function. Social support liminary observation is that although health-related behaviors, researchers might gain greater specificity and prediction by ex- stressful events, and depression clearly influence physiological amining specific types of social relationships. The studies sum- processes in their own right, they do not appear to be major marized in this review whose researchers examined cross-cul- pathways by which social support influences physiological func- tural and gender effects of social support are examples of such tion. Future researchers might attempt to characterize the po- applications. In addition, behavioral data obtained during lab- tential psychological mechanisms responsible for the associa- oratory studies may amplify the relationships found between tions between social support and physiological processes based self-report data and physiological processes (e.g., Kiecolt- on the larger social support literature (see S. Cohen, 1988 ). Glaser et al., 1993; Malarkey, Kiecolt-Glaser, Pearl, & Glaser, Prior research on social support processes has been focused 1994). on support as a situational variable, however, social support Most of the studies reviewed in this article have conceptual- may also be conceptualized as a stable individual differences ized social support as a unidimensional construct. As noted ear- variable (I. G. Sarason et al., 1986 ). It is therefore possible that lier, multidimensional assessments may allow for an examina- measures of social support may be influenced by personality tion of more specific associations and mechanisms (Uchino, factors that are correlated with social relationships. As a result, Cacioppo, Malarkey, Glaser, & Kiecolt-Glaser, 1995). For in- social support may be associated with physiological function stance, Seeman et al. (1994) found that emotional support was through personality processes that influence perceptions of sup- a more consistent predictor of neuroendocrine function than port (Bolger & Eckenrode, 1991 ). However, consistent with re- informational support. More important, such specificity would search indicating that social support predicts objective health have been lost if an aggregate measure of social support was outcomes, even after statistically controlling for personality fac- used. The relative importance of specific dimensions of social tors (House et al., 1988), social support appears to predict physiological function even after statistically controlling for the effects of personality variables or their confounding mechanism (e.g., Kiecolt-Glaser et al., 1991; Uchino et al., 1992; Uchino, data on combined tests of significance for the relationships that they Kiecolt-Glaser, & Cacioppo, 1994). Note that personality fac- examined. Therefore, some of the relationships between social support tors such as neuroticism and extraversion appear to have a may have been statistically significant, albeit characterized by small to larger effect on more subjective measures of health, such as self- moderate effect sizes. Second, the aggregation of such diverse measures reported symptomology (Watson & Pennebaker, 1989). How- as indices of physical health and stress outcomes (see above) may obscure reliable relationships within particular measures (e.g., blood pres- ever, the personality dimension of trait hostility is related to ob- sure and catecholamines). Finally, and perhaps most important, the jective measures of health (Barefoot, Dahlstrom, & Williams, selection of studies in the meta-analysis was limited. None of the major 1983; Shekelle, Gale, Ostfeld, & Paul, 1983) and appears to prospective studies on social relationship and mortality were included have an important interpersonal component (T. W. Smith & in the meta-analysis (see House et al., 1988, for a review). In addition, Frohm, 1985; T. W. Smith, Pope, Sanders, AUred, & O'Keeffe, there appears to be little overlap in the studies examined in Smith et al. 1988). Future researchers could profitably examine the in- and the present review. The reasons for this discrepancy is unclear be- terface between social relationships and personologic factors, cause similar databases were searched. Unfortunately, Smith et al. did especially in reference to physical health (McGonigle, Smith, not report the keywords used for their search. We reviewed 81 studies Benjamin, & Turner, 1993; T. W. Smith, in press). focusing on the physiological mechanisms that may be underlying the In the present review, we have focused on the positive aspects association between social support and health, whereas the Smith et al. of social relationships. There is a growing body of literature in- review only included 61 studies, but they also examined other diverse outcomes (i.e., psychological, stress-related, and physical outcomes). dicating that negative aspects of social relationships are inde- Therefore, it appears that our review on the physiological processes un- pendent of positive aspects of support (Ruehlman & Karoly, derlying the associations between social support and health was more 1991 ) and an important predictor of psychological functioning comprehensive than the Smith et al. review. (Pagel, Erdly, & Becker, 1987; Rook, 1984; Rook & Pietromo- SOCIAL SUPPORT AND PHYSIOLOGY 523 naco, 1987). In addition, in their meta-analysis Herbert and ski et al., 1988). For instance, Krantz et al. ( 1991 ) examined Cohen (1993a) suggest that such negative aspects of social rela- stress-induced myocardial ischemia in coronary disease pa- tionships may have important effects on immune function. tients. Results revealed that patients with more severe myocar- Therefore, the assessment of both positive and negative aspects dial ischemia during mental stress were characterized by greater of social relationship by future researchers might prove worth- SBP reactivity (also see Rozanski et al., 1988). These data are while. In a recent study demonstrating the use of examining consistent with research indicating that social support also pre- both positive and negative aspects of social relationships, LOP- dicts survival after the diagnosis of cardiovascular disorders ore (1992) found that conflict with roommates or friends was a (e.g., Berkman, Leo-Summers, & Horwitz, 1992). significant predictor of psychological distress. More important, Activation ofa sympathetic-adrenergic mechanism may also positive aspects of support served to attenuate this association. influence immune system functioning. Direct sympathetic in- The separability of positive and negative components is an im- nervation exists for both primary and secondary lymphoid or- portant and emerging conceptual theme in several areas of psy- gans (see review by Felten & Felten, 1991 ), and lymphocytes chological inquiry (Cacioppo & Berntson, 1994; Kiecolt-Glaser have beta-adrenergic receptors (Khan, Sansoni, Silverman, En- et al., 1993; Watson, Clark, & Tellegen, 1988 ). gleman, & Melmon, 1986; Williams, Snyderman, & Lefkowitz, An important issue in the present review is an examination 1976). More significantly, catecholamines appear to influence of the psychometric properties of the physiological assessments aspects of the cellular and humoral immune response (Madden because they bear on the potential mechanisms linking social & Livnat, 1991; Sanders & Munson, 1985). In addition, labo- support to long-term physical health. Measurement reliability ratory studies suggest that short-term alterations in immune tends to be high in the assessment of cardiovascular, endocrine, function to acute psychological stress are largest in individuals and immune function. Furthermore, prior research has docu- showing greater cardiac-sympathetic reactivity (Cacioppo et mented the temporal stability of the cardiovascular assessments al., 1995). (e.g., Manuck, 1994). However, little data currently exist on It is important to separate the short-term and longer term the temporal stability of assessments of endocrine and immune effects of sympathetic activation because they may have differ- function (cf. Fletcher et al., in press; Marsland et al., 1995). In ent implications for the effects of social support on the etiology this regard, the lack of prospective studies examining the link of disease progression. Although the data on this issue are lim- between social support and physiological function is notewor- ited, chronic stress may be associated with increased sympa- thy. Prospective studies may prove helpful in examining the thetic activity (Baum, 1990; Irwin et al., 1991 ) and a down reg- temporal stability and association between social support, phys- ulation of beta-adrenergic receptors (Dimsdale, Mills, Patter- iological processes, and potential mechanisms over appreciable son, Ziegler, & Dillon, 1994). Such changes appear to be periods of time. associated with decrements in immune function (Irwin et al., The effects of social support appear most reliable on blood 1991 ). For instance, Irwin et al. ( 1991 ) found elevated levels of pressure, catecholamines, and aspects of both the cellular and a sympathetic neurotransmitter, neuropeptide Y, in caregivers humoral immune response. The laboratory reactivity studies of patients with Alzbeimer's disease compared with matched with young individuals suggest that social support may lower controls and reported that levels of neuropoptide Y were nega- cardiovascular reactivity to psychosocial stressors (e.g., Gerin tively related to NK cell lysis. Given the long-term nature of et al., 1992; Lepore et al., 1993). Cardiovascular reactivity to chronic stress, the repeated activation of a sympathetic-adren- acute psychological stressors is thought to reflect, in part, sym- ergic mechanism may have relatively long-term effects on the pathetic-adrenergic activation (Light, 1981 ). In addition, immune system. exposure to acute psychological stressors appears to activate the In addition to the evidence indicating the importance ofsym- sympathetic adrenal medullary (SAM) system as evidenced by pathetic-adrenergic influences, glucocorticoids have important increased plasma catecholamines levels (Cacioppo et al., 1995; effects on many aspects of immune function (Cupps & Fauci, Manuck, Cohen, Rabin, Muldoon, & Bachen, 1991; Sgoutas- 1982). As reviewed earlier, there are important reasons why Emch et al., 1994). Therefore, one model of these data suggests prior studies may not have found a relationship between social that social support may be related to long-term health through support and tonic cortisol levels. In addition, cortisol reactivity a common sympathetic-adrenergic mechanism. in the form of "acute bursts" may be important in explaining We also reviewed evidence on the association between social variance between social support and health. Consistent with support and tonic blood pressure levels. These studies tested this reasoning, preliminary data from our laboratory suggest middle-aged to older individuals and suggested that social sup- that cortisol reactivity to acute psychological stress predicted port was associated with lower blood pressure levels. As noted older adult reactions to an influenza vaccination (i.e., influenza earlier, heightened cardiovascular reactivity in the young may virus induced interleukin-2 levels in vitro) given 7 to 8 months translate to gradual elevations in tonic blood pressure levels earlier (see Cacioppo, 1994). To date, little idiographic data ex- over the years (e.g., Light et al., 1992). This developmental se- ist on the influence of social support on cortisol reactivity to quence involving mediation by a sympathetic-adrenergic acute psychosocial stressors. In addition, an examination of re- mechanism is consistent with animal models of hypertension ceptor functioning may prove helpful in obtaining an integrated (Folkow et al., 1973; Hallback & Folkow, 1974). understanding of not only the association between social sup- Sympathetic-adrenergic activation may not only have im- port and hormones of the hypothalamic pituitary-adrenal axis plications for the development of cardiovascular disorders but (Seeman & Robbins, 1994) but also hormones of the SAM may also prime significant coronary events in diseased popula- (Mills & Dimsdale, 1993). tions (Kamarck & Jennings, 1991; Krantz et al., 1991; Rozan- Future research will help to understand the potential impor- 524 UCHINO, CACIOPPO, AND KIECOLT-GLASER tance of examining specific patterns of sympathetic-adrenergic sponses (Kasprowicz, Manuck, Malkoff, & Krantz, 1990; activity (Cacioppo et al., 1992 ). For instance, research on neuro- Sherwood, Dolan, & Light, 1990). A specifiable subset of in- endocrine-immune interactions during acute psychological stress dividuals achieve blood pressure changes primarily through suggests that cardiac-sympathetic reactivity, as assessed by pre- vascular mechanisms (i.e., peripheral resistance), whereas a ejection period, appears to be closely linked to activation of the different subset of individuals achieve comparable blood HPA, whereas catecholamines do not appear to operate through pressure changes primarily through hemodynamic mecha- such a mechanism (Cacioppo et al., 1995; Manuck, Cohen, nisms (i.e., cardiac output). The impedance cardiograph Rabin, Muldoon, & Bachen, 1991; Sgoutas-Emch et al., 1994; (see Sherwood, Allen, et al., 1990) provides noninvasive esti- Uchino, Cacioppo, Malarkey, & Glaser, 1995). Research designs mates of cardiac output and total peripheral resistance and that incorporate assessments across the cardiovascular, endocrine, should prove useful in future investigations of the underlying and immune systems may help elucidate more specific pathways basis of both tonic and phasic blood pressure. responsible for the effects of social support across physiological We have focused on the cardiovascular, endocrine, and im- processes. To this point, it should be noted that of the 81 studies mune systems as potential windows through which to examine reviewed in this article, only 5 obtained simultaneous measures the long-term health consequence of social relationships. The from two of the different physiological systems (e.g., cardiovascu- intervention studies examining hypertensive patients provide lar and endocrine) and that none of the studies obtained measures data to show how social support could have important health across all three systems. It is also noteworthy that there is a sur- consequences. Animal models have demonstrated the predic- prising dearth of research on social support and neuroendocrine tive utility of cardiovascular reactivity in the development of processes in humans. The neuroendocrine system almost cer- cardiovascular disorders (Folkow et al., 1973; Hallback & Fol- tainly serves as an important gateway between personal relation- kow, 1974; Manuck et al., 1983). Less is known, however, about ships and health; even daily psychological stressors can provoke the long-term effects of heightened cardiovascular reactivity in the release of pituitary and sympathetic adrenomeduUary hor- humans, although supportive evidence is mounting (e.g., Light mones that have multiple effects, including alterations in the car- et al., 1992; Manuck, 1994). Research on social support in at- diovascular and immune systems (e.g., Cacioppo et al., 1995; Ma- risk populations and continued theoretical development in the larkey et al., 1994). reactivity literature may help to clarify the link between cardio- We should note that unidimensional conceptualizations of vascular function and physical health. physiological processes may be at least as significant an ob- Similarly, animal models have demonstrated the importance stacle to illuminating the mechanisms underlying the health of alterations in the endocrine and immune systems on infec- benefits of social support, as are unidimensional conceptual- tions and tumor growth (Habu, Akamatsu, Tamaoki, & Oku- izations of social support. Cardiovascular reactivity, for in- mura, 1984; Lewis et al., 1983; Sheridan, Feng, Bonneau, Ma- stance, has tended to be treated as a unidimensional (and oc- larkey, & Hughes, 1991 ). In addition, some studies have linked casionally as a unidirectional) construct ranging from low to alterations in immune function with various health conse- high, reflecting individual differences in adrenergic reactivity quences in humans (Cogen, Stevens, Cohen-Cole, Kirk, & Free- to daily stressors and behavioral challenges. Although adren- man, 1982; Fletcher, Baron, Ashman, Fischl, & Klimas, 1987; ergic activity exerts predominant control over the vasculature Kiecolt-Glaser & Glaser, 1995; Lumio, Welin, Hirvonen, & We- and cardiac inotropy, cardiac chronotropy (and, thus, ber, 1983; Murasko, Weiner, & Kaye, 1988), including cancer, cardiac output) is a joint function of sympathetic and vagal infectious illness, and HIV progression. Future studies on social activity. Moreover, vagal as well as sympathetic influences on support and physiological processes that include measures of the heart are evident both at rest and in response to daily physical health status would be helpful in clarifying the links to tasks, challenges, and stressors (e.g., M. T. Allen & Crowell, actual health outcomes. 1989; Grossman, Stemmler, & Meinhardt, 1990; see reviews As noted by House et al. (1988), the effects of social support by Cacioppo, 1994; and Porges, 1992), and the vagal and on health may be particularly important for older adults: sympathetic outflows to the heart can vary reciprocally, non- reciprocally (e.g., coactivation), or independently (Berntson, Changes in marital and childbearing patterns and in the age struc- Cacioppo, & Quigley, 1991, 1993; Cacioppo, Uchino, & ture of our society will produce in the 21st century a steady increase of the number of older people who lack spouses or chil- Bernstson, 1994). An individual's classification as high in dren-the people to whom older people often turn for relatedness cardiovascular reactivity ignores possible individual differ- and support. Thus, just as we discover the importance of social ences in the autonomic origins of this reactivity. Variations relationships for health, and see an increasing need for them, their in the autonomic origins of cardiovascular reactivity have prevalence and availability may be declining. (p. 544 ) generally been relegated to the error term, a practice that has obscured relationships between autonomic, endocrine, and Chronological aging is typically associated with changes in the immune responses and may obscure relationships between cardiovascular (Uchino et al., 1992) and endocrine systems autonomic processes and social, behavioral, or health out- (Meites, Goya, & Takahashi, 1987), a down regulation of the comes. immune system (Burns & Goodwin, 1990), and declines in As the case for heart rate, similar conceptual issues relate physical health (Effros & Walford, 1987; Kart, Metress, & Me- to the assessment of blood pressure. Blood pressure is a func- tress, 1992). Therefore, alterations in these compromised phys- tion of both cardiac output and peripheral resistance. Recent iological systems may have significant health consequences in research has demonstrated considerable individual differ- older populations. More important, the studies summarized in ences in the underlying mechanisms of blood pressure re- this review are consistent with the notion that social support SOCIAL SUPPORT AND PHYSIOLOGY 525 may moderate such physiological processes in older adults. sents only the first wave of research describing associations be- Given the importance of close affective social bonds in these tween social processes and health. adults (Antonucci & Akiyama, 1987; Carstensen, 1992) and impending demographics shifts that may curtain the availability of familial sources of support, facilitating the development References of close emotional bonds among them may be particularly Ader, R., Felton, D. L., & Cohen, N. ( 1991 ). Psychoneuroimmunology important. (2nd ed.). New York: Academic Press. The studies summarized in this review may have relevance Akerstedt, T., & Levi, L. (1978). Circadian rhythms in the secretion of for the aging process more generally. Research on chronological cortisol, adrenaline, and noradernaline. European Journal of Clinical age and physiological function suggests that such physiological Investtgations, 8, 57-58. changes are not a biological invariant with aging (E. L. Smith, Allen, K. M., Blascovich, J., Tomaka, J., & Kelsey, R. M. ( 1991 ). Pres- 1984) and that social factors may play a role in the aging process ence of human friends and pet dogs as moderators of autonomic responses to stress in women. Journal of Personality and Social Psy- (Szklo, 1979; Uchino et al., 1992). The results of this review chology 61, 582-589. suggest that social support has beneficial effects on physiological Allen, M. T., & Crowell, M. D. (1989). Patterns of autonomic response processes across different age groups. The net effect of such pro- during laboratory stress. Ps)~:hophysiology, 26, 603-614. cesses may be to biologically age the individual at a slower rate. Andersson, L. ( 1985 ). Intervention against loneliness in a group of el- Consistent with this hypothesis, we have found that social sup- derly women: An impact evaluation. Social Science and Medicine, port predicts age-related differences in blood pressure. More 20, 355-364. specifically, individuals low in social support were characterized Antonucci, T. C., & Akiyama, H. (1987). Social networks in adult life by age-related increases in blood pressure, whereas individuals and a preliminary examination of the convoy model. Journal of Ger- high in social support were characterized by low and compara- ontology, 42, 519-527. ble blood pressure across the ages (Uchino et al., 1992; Uchino, Arnetz, B. B., Edgren, B., Levi, L., & Otto, U. ( 1985 ). Behavioural and endocrine reactions in boys scoring high on Sennton neurotic scale Cacioppo, Malarkey, Glaser, & Kiecolt-Glaser, 1995). viewing an exciting and partly violent movie and the importance of In conclusion, understanding what precisely are the relation- social support. Social Science and Medicine, 20, 731-736. ships between social support and health outcomes in specific Arnetz, B. B., Theorell, T., Levi, L., Kallner, A., & Eneroth, P. (1983). populations will likely change as refinements are made in the An experimental study of social isolation of elderly people: Psychoen- conceptualization and assessment of theoretical constructs, docrine and metabolic effects. Psychosomatic Medicine, 45, 395-406. such as social support, construals and coping responses, behav- Arnetz, B. B., Wasserman, J., Petrini, B., Brenner, S., Levi, L., Eneroth, iorally relevant physiological events, and so on. As research P., Salovaara, H., Hjelm, R., Salovaara, L., Theorell, T., & Petterson, moves from a description of associations to a delineation of un- I. ( 1987 ). Immune function in unemployed women. Psychosomatic derlying mechanisms, questions of causative factors take on Medicine, 49, 3-12. added importance. Considerable research exists, for instance, Barefoot, J. C., Dahlstrom, W. G., & Williams, R. B. (1983). Hostility, CHD incidence, and total mortality: A 25 year follow-up study of 255 demonstrating that neurochemical events can influence social physicians. Psychosomatic Medicine, 45, 59-63. processes and that social processes influence neurochemical Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator dis- events. Studies of mating behavior in the ring-necked dove indi- tinction in social psychological research: Conceptual, strategic, and cate that social behaviors (e.g., male strutting and cooing) trig- statistical considerations. Journal of Personality and Social Psychol- ger hormonal changes (e.g., increased production of estrogen in ogy 51, 1173-1182. the female dove), which predispose the female toward a new set Baron, R. S., Cutrona, C. E., Hicklin, D., Russell, D. W., Lubaroff, of behaviors (e.g., courtship and copulation), which results in D. M. (1990). Social support and immune function among spouses additional reciprocal influences between hormones and social of cancer patients. Journal of Personality and Social Psychology, 59, behaviors until the newborns are reared (Erikson & Lehrman, 344-352. 1964; Lehrman, 1964; Martinez-Vargas & Erikson, 1973). Re- Barrera, M. (1986). Distinctions between social support concepts, measures, and models. American Journal of Community Psychology, ciprocal influences between social and physiological events have 14, 413-445. been found in primates, as well. Testosterone levels in male pri- Baum, A. (1990). Stress, intrusive imagery, and chronic distress. mates, for example, have been found to promote sexual behav- Health Psychology, 9, 653-675. ior, whereas the availability of sexually receptive females, in Baum, A., & Grunberg, N. (1995). Measurement of stress hormones. turn, has been found to influence testosterone levels in male In S. Cohen, R. C. Kessler, & L. U. Gordon (Eds.), Measuring stress: primates (Bernstein, Gordon, & Rose, 1983; Bernstein, Rose, A guide for health and social scientists (pp. 175-192). New York: & Gordon, 1974; Rose, Gordon, & Bernstein, 1972; see Caci- Oxford University Press. oppo & Berntson, 1992). Indeed, understanding the function of Berkman, L. E, Leo-Summers, L., & Horwitz, R. I. ( 1992 ). Emotional hormones may have been far more rudimentary if not for anal- support and survival after myocardial infarction. Annals of Internal Yses of their effects on social behavior and for the effects of social Medicine, 117, 1003-1009. Bernstein, I. S., Gordon, T. P., & Rose, R. M. (1983). The interaction behavior on hormonal changes. There are, of course, physiolog- of hormones, behavior, and social context in nonhuman primates, tn ical mechanisms underlying these phenomena, but the identi- B. B. Svare (Ed.), Hormones and aggressive behavior (pp. 535-561 ). fication and understanding of these mechanisms are better New York: Plenum Press. served by systematic investigations within and across multiple Bernstein, I. S., Rose, R. M., & Gordon, T. P. (1974). Behavioral and levels of analysis rather than by an exclusive focus on any one environmental events influencing primate testosterone levels. level of analysis. Thus, the research reviewed here specifies a set Journal of Human Evolution, 3, 517-525. of associations in need of mechanistic explanations but repre- Berntson, G. G., Cacioppo, J. T., & Quigley, K. S. ( 1991 ). Autonomic 526 UCHINO, CACIOPPO, AND KIECOLT-GLASER

determinism: The modes of autonomic control, the doctrine of auto- individual differences. Journal of Personality and Social Psychology, nomic space, and the laws of autonomic constraint. PsychologicalRe- 62, 110-128. view, 98, 459-487. Calabrese, J. R., Kling, M. A., & Gold, P. W. (1987). Alterations in Berntson, G. G., Cacioppo, J. T., & Quigley, K. S. (1993). Cardiac psy- immunocompetence during stress, bereavement, and depression: Fo- chophysiology and autonomic space in humans: Empirical perspec- cus on neuroendocrine regulation. American Journal of Psychiatry, tives and conceptual implications. Psychological Bulletin, 114, 296- 144, 1123-1134. 322. Carr, D. J. ( 1991 ). The role ofendogeneous opioids and their receptors Bissette, G., & Ritchie, J. C. (1992). Radioimmunoassay methods. In in the immune system. Proceedings of the Society of Experimental C. B. Nemeroff (Ed.), Neuroendocrinology (pp. 39-50). London: Biology and Medicine, 198, 710-720. CRC Press. Cartensen, L. L. (1992). Social and emotional patterns in adulthood: Bland, S. H., Krogh, V., Winkelstein, W., & Trevisan, M. ( 1991 ). Social Support for socioemotional selectivity theory. Psychology and Aging, network and blood pressure: A population study. Psychosomatic 7, 331-338. Medicine, 53, 598-607. Cassell, J. (1976). The contribution of the social environment to host Blalock, E. J. (1989). A molecular basis for bidirectional communica- resistance. American Journal of Epidimiology, 104, 107-123. tion between the immune and neuroendocrine systems. Physiological Chard, T. (1982). An introduction to radioimmunoassay and related Reviews, 69, 1-31. techniques. In T. S. Work & E. Work (Eds.), Laboratory techniques in Blazer, D. G. (1982). Social support and mortality in an elderly com- biochemistry and molecular biology ( pp. 1-284). New York: Elsevier munity population. American Journal of Epidemiology, 115, 684- Biomedical Press. 694, Clifford, P. A., Tan, S., & Gorsuch, R. L. (1991). Efficacy of a self- Bolger, N., & Eckenrode, J. (1991). Social relationships, personality, directed behavioral health change program: Weight, body composi- and anxiety during a major stressful event. Journal ofPersonality and tion, cardiovascular fitness, blood pressure, health risk, and psycho- Social Psychology, 61,440-449. social mediating variables. Journal of Behavioral Medicine, 14, 303- Borysenko, M. (1987). The immune system: An overview. Annals of 323. Behavioral Medicine, 9, 3-10. Cobb, S. (1974). Physiologic changes in men whose jobs were abol- Boyce, T. W., & Chesterman, E. (1990). Life events, social support, and ished. Journal of Psychosomatic Research, 18, 246-258. cardiovascular reactivity in adolescence. Developmental and Behav- Cobb, S. (1976). Social support as a moderator of life stress. Psychoso- ioral Pediatrics, 11, 105-111. matic Medicine, 38, 300-314. Broadhead, W. E., Kaplan, B. H., James, S. A., Wagner, E. H., Schoen- Coe, C. L. (1993). Psychosocial factors and immunity in nonhuman bach, V. J., Grimson, R., Heyden, S., Tibblin, G., & Gehlbach, S. H, primates: A review. Psychosomatic Medicine, 55, 298-308. ( 1983 ). The epidemiologic evidence for a relationship between social Cogen, R. B., Stevens, A. W., Cohen-Cole, S., Kirk, K., & Freeman, support and health. American Journal of Epidemiology, 117, 521- A. (1982). Leukocyte function in the etiology of acute necrotizing 537. ulcerative gingivitis. Journal of Periodontology, 54, 402-407. Burns, E. A., & Goodwin, J. S. (1990). Immunology and infectious Cohen, J., & Cohen, P. (1983). Applied multiple regression~correlation disease. In C. K. Cassel, D. E. Risenberg, C. B. Sorensen, & J. R, analysis for the behavioral sciences. Hillsdale, N J: Edbaum. Walsh (Eds.), Geriatric medicine (pp. 312-329). New York: Cohen, S. (1988). Psychosocial models of the role of social support in Springer-Verlag. the etiology of physical disease. Health Psychology, 7, 269-297. Cacioppo, J. T. (1994). Social neuroscience: Autonomic, neuroendo- Cohen, S., & McKay, G. (1984). Social support, stress and the buffering crine, and immune responses to stress. Psychophysiology, 31, 113- hypothesis: A theoretical analysis. In A. Baum, S. E. Taylor, & J. E. 128. Singer (Eds.), Handbook of psychology and health (pp. 253-267). Cacioppo, J. T., & Berntson, G. G. ( 1992 ). Social psychological contri- Hillsdale, N J: Erlbaum. butions to the decade of the brain: Doctrine of multilevel analysis, Cohen, S., & Syme, S. L. (1985), Socialsupport andhealth. New York: American Psychologist, 47, 1019-1028. Academic Press. Cacioppo, J. T., & Berntson, G. G. (1994). The relationship between Cohen, S., & Wills, T. A. (1985). Stress, social support, and the buff- attitudes and evaluative space: A critical review, with emphasis on the ering hypothesis. Psychological Bulletin, 98, 310-357. separability of positive and negative substrates. Psychological Bulle- Collins, N. L., Dunkel-Schetter, C., Lobel, M., & Scrimshaw, S. C. M. tin, 115, 401-423. ( 1993 ). Social support in pregnancy: Psychosocial correlates of birth Cacioppo, J. T., Berntson, G. G., & Andersen, B. L. (1991). Psycho- outcomes and postpartum depression. Journal of Personality and So- physiological approaches to the evaluation of psychotherapeutic pro- cial Psychology, 65, 1243-1258. cess and outcome, 1991: Contributions from social psychophysiology. Cottington, E. M., Brock, B. M., House, J. S., & Hawthorne, V. M. Psychological Assessment, 3, 321-336. (1985). Psychosocial factors and blood pressure in the Michigan Cacioppo, J. T., Malarkey, W. B., Kiecolt-Glaser, K. G., Uchino, B. N., statewide blood pressure study. American Journal of Epidemiology, Sgoutas-Emch, S. A., Sheridan, J. F., Berntson, G. G., & Glaser, R. 121.515-529. ( 1995 ). Heterogeneity in neuroendocrine and immune responses to Crary, B., Borysenko, M., Sutherland, D. C., Kutz, I., Borysenko, J. Z., brief psychological stressors as a function of autonomic cardiac acti- & Benson, H. (1983). Decrease in mitogen responsiveness of mono- vation. Psychosomatic Medicine, 57, 154-164. nuclear cells from peripheral blood after epinephrine administrations Cacioppo, J. T., Uchino, B. N., & Berntson, G. G. (1994). Individual in humans. Journal of lmmunology, 130, 694-697. differences in the autonomic origins of heart rate reactivity: The psy- Cupps, T. R., & Fauci, A. S. ( 1982 ). Corticosteroid-mediated immuno- chometrics of respiratory sinus arrhythmia and preejection period. regulation in man. Immunological Review, 65, 133-155. Psychophysiology, 31, 412-419. Cutrona, C. E., & Russell, D. W. (1990). Type of social support and Cacioppo, J. T., Uchino, B. N., Crites, S. L., Snydersmith, M. A., Smith, specific stress: Toward a theory of optimal matching. In B. R. Sarason, G., Berntson, G. G., & Lang, P. J. (1992). The relationship between I. G. Sarason, & G. R. Pierce (Eds.), Social support: An interactional facial expressiveness and sympathetic activation in emotion: A criti- view (pp. 319-366). New York: Wiley. cal review, with emphasis on modeling underlying mechanisms and Dimsdale, J. E., Mills, P., Patterson, T., Ziegler, M., & Dillon, E. (1994). SOCIAL SUPPORT AND PHYSIOLOGY 527

Effects of chronic stress on beta-adrenergic receptors in the homeless. Effects of calcium entry blockers and calmodulin antagonists. Ste- Psychosomatic Medicine, 56, 290-295. roids, 49, 601-616. Dressier, W. W. (1980). Coping dispositions, social supports, and health Gelman, R., Cheng, S.-C., Kidd, P., Waxdal, M., & Kagan, J. (1993). status. Ethos, 8, ! 46-171. Assessment of the effects of instrumentation, monoclonal antibody, Dressier, W. W. (1983). Blood pressure, relative weight, and psychoso- and fluorochrome on flow cytometric immunophenotyping: A report cial resources. Psychosomatic Medicine, 45, 527-536. based on 2 years of the NIAID DAIDS flow cytometry quality assess- Dressier, W. W. ( 1991 ). Social support, lifestyle incongruity, and arte- ment program. Clinical Immunology and Immunopathology, 66, rial blood pressure in a southern Black community. Psychosomatic 150-162. Medicine, 53, 608-620. Gerin, W., Milner, D., Chawla, S., & Pickering, T. G. (1995). Social Dressier, W. W., Grell, G. A., Gallagher, P. N., & Viteri, E E. (1992). support as a moderator of cardiovascular reactivity in women: A test Social factors mediating social class differences in blood pressure in a of the direct effects and buffering hypotheses. Psychosomatic Medi- Jamaican community. Social Science and Medicine, 35, 1233-1244. cine. 57, 16-22. Dressier, W. W., Mata, A., Chavez, A., Viteri, E E., & Gallagher, P. Gerin, W., Pieper, C., Levy, R., & Pickering, T. G. (1992), Social sup- (1986). Social support and arterial blood pressure in a central Mexi- port in social interaction: A moderator of cardiovascular reactivity. can community. Psychosomatic Medicine, 48, 338-350. Psychosomatic Medicine, 54, 324-336. Dressier, W. W., Santos, J. E., & Viteri, E E. (1986). Blood pressure, Gill, A. A., Veigl, V. L., Shuster, J. J., & Notelovitz, M. (1984). A well ethnicity, and psychosocial resources. Psychosomatic Medicine, 48, woman's health maintenance study comparing physical fitness and 509-519. group support programs. Occupational Therapy Journal ofResearch, Earp, K. L., Ory, M. G., & Strogatz, D. S. (1982). The effects of family 4, 286-308. involvement and practitioner home visits on the control of hyperten- Glaser, R., Kiecolt-Glaser, J. K., Bonneau, R., Malarkey, W., & Hughes, sion.American Journal of Public Health, 72, t 146-1 t54. J. (1992). Stress-induced modulation of the immune response to re- Edens, J. L., Larkin, K. T., & Abel, J. L. (1992). The effects of social combinant hepatitis B vaccine. Psychosomatic Medicine, 54, 22-29. support and physical touch on cardiovascular reactions to mental Goidl, E. A. ( 1987 ). Aging and the immune response. New York: Marcel stress. Journal of Psychosomatic Research, 36, 371-382. Dekker. Effros, R. B., & Walford, R. L. (1987). Infection and immunity in rela- Goodkin, K., Blaney, N. T., Feaster, D., Fletcher, M., Baum, M. K., tion to aging. In E, A. Goidl (Ed.), Aging and the immune response Mantero-Atienza, E., Klimas, N. G., Millon, C., Szapocznik, J., & (pp. 45-65 ). New York: Marcel Dekker. Eisdorfer, C. (1992). Active coping style is associated with natural Ely, D. L., & Mostardi, R. A. (1986). The effects of recent life events killer cell cytotoxicity in asymptomatic HIV- 1 seropositive homosex- stress, life assets, and temperament pattern on cardiovascular risk fac- ual men. Journal of Psychosomatic Research, 36, 635-650. tors for Akron city police officers. Journal of Human Stress, 12, 77- Goodwin, J. S., Searles, R. P., & Tung, K. S. K. (1982). Immunological 91. responses of a healthy elderly population. Clinical and Experimental Erfurt, J. C., Foote, A., & Heirich, M. A. ( 1991 ). Worksite wellness Immunology, 48, 403-410. program: Incremental comparison of screening and referral alone, Gorlin, R. ( 1976 ). Coronary artery disease. London: Saunders. health education, follow-up counseling, and plant organization. Greenspan, E S., & Baxter, J. D. (1994). Basic and clinical endocrinol- American Journal of Health Promotion, 5, 438--448. ogy Norwalk, CI': Appleton & Lange. Erikson, C., & Lehrman, D. (1964). Effects of castration of male ring Groer, M. W., Humenick, S., & Hill, P. D. (1994). Characterizations doves upon ovarian activity of females. Journal of Comparative and and psychoneuroimmunologic implications of secretory immuno- Physiological Psychology, 58, 164-166. globulin A and cortisol in preterm and term breast milk. Journal of Esterling, B. A., Kiecolt-Glaser, J. K., Bodnar, J. C., & Glaser, R. Perinatal and Neonatal Nursing, 7, 42-51. (1994). Chronic stress, social support, and persistent alterations in Grossman, P., Stemmler, G., & Meinhardt, E. (1990). Paced respira- the natural killer cell response to cytokines in older adults. Health tory sinus arrythmia as an index of cardiac parasympathetic tone Psychology 13, 291-299. during varying behavioral tasks. Psychophysiology, 27, 404-416. Felton, S. Y., & Felten, D. L. ( 1991 ). lnnervation of lymphoid tissue. In Gunnar, M. R. (1992). Reactivity of the hypothalamic-pituitary-adre- R. Ader, D. L. Felton, & N. Cohen (Eds.), Psychoneuroimmunology nocortical system to stressors in normal infants and children. Pediat- (pp. 27-69). New York: Academic Press. rics, 90, 491-497. Fleming, R., Baum, A., Gisriel, M. M., & Gatchel, R. J. (1982). Medi- Habu, S., Akamatsu, K., Tamaoki, N., & Okumura, K. (1984), In vivo ating influences of social support on stress at Three Mite Island. significance of NK cells on resistance against virus (HSV-1 ) infec- Journal of Human Stress, 8. 14-22. tions in mice. Journal of Immunology, 133, 2743-2747. Fletcher, M. A., Baron, G. C., Ashman, M. R., Fischl, M. A., & Klimas, Hallback, M., & Folkow, B. (1974). Cardiovascular responses to acute N. G. (1987). Use of whole blood methods in assessment of immune mental "stress" in spontaneously hypertensive rats. Acta Physiologica parameters in immunodeficiency states. Diagnostic Clinical Immu- Scandinavica, 90, 684-693. nology, 5, 69-81. Hansell, S. (1985). Adolescent friendship networks and distress in Fletcher, M. A., Klimas, N. G., Morgan, R., & Gjerset, G. (in press). school. Social Forces, 63, 698-715. Lymphocyte proliferation. In N. Rose (Ed.), Manualofclinical labo- Hanson, B. S., Isacsson, S. O., Janzon, L., Lindell, S. E., & Rastam, ratory immunology Washington, DC: American Association for L. (1988). Social anchorage and blood pressure in elderly men--A Microbiology. population study. Journal of Hypertension, 6, 503-5 I0. Folkow, B., Hallback, M., Lundgren, Y., Sivertsson, R., & Weiss, L. Heitzmann, C. A., & Kaplan, R. M. (1988). Assessment of methods for (1973). Importance of adaptive changes in vascular design for estab- measuring social support. Health Psychology, 7, 75-109. lishment of primary hypertension, studied in man and in spontane- Henderson, S., Byrne, D. G., Duncan-Jones, P., Scott, R., & Adcock, ously hypertensive rats. Circulation Research, 32-33 (Suppl, 1 ), 2- S. (1980). Social relationships, adversity, and neurosis: A study of 13. associations in a general population sample. British Journal of Psy- Gatti, G., Masera, R., Cavallo, R., Sartori, M. L., Delponte, D., Carig- chiatry, 136, 574-583. nola, R., Salvadori, A., & Angeli, A. (1987). Studies on the mecha- Herbert, T. B., & Cohen, S. (1993a). Depression and immunity: A nism of cortisol inhibition of human natural killer cell activity: meta-analytic review. Psychological Bulletin, 113, 472-486. 528 UCHINO, CACIOPPO, AND KIECOLT-GLASER

Herbert, T. B., & Cohen, S. (1993b). Stress and immunity in humans: effects on cardiovascular functioning. Psychotherapy and Psychoso- A meta-analytic review. Psychosomatic Medicine, 55, 364-379. matics, 34, 88-109. Herrera, J. A., Alvarado, J. P, & Martinez, J. E. (1988). The psychoso- Kasprowicz, A. L., Manuck, S. B., Malkoff, S. B., & Krantz, D. S. cial environment and the cellular immunity in the pregnant patient. (1990). Individual differences in behaviorally evoked cardiovascular Stress Medicine, 4.49-56. response: Temporal stability and hemodynamic patterning. Psycho- HolT, P. S, & Lowenstein, A. ,l. (1994). Professional support: An essen- physiology, 27, 605-619. tial component in a cardiovascular risk reduction program. Journal Kaufmann, G. M., & Beehr, T. A. (1986). Interactions between job of Gerontological Nursing, 29, 28-32. stressors and social support: Some counterintuitive results. Journal Houben, G. ,l., Diederiks, J. P. M., Kant, Y., & Notermans, ,l. V. H. ofApplied Psychology, 71, 522-526. (1990). Rationalization in garages in the Netherlands and its effects Khan, M. M., Sansoni, P., Silverman, E. D., Engleman, E. G., & Mel- on occupational health. Work andStress, 4, 179-189. mon, K. L. (1986). Beta-adrenergic receptors on human suppressor, House, J. S., Landis, K. R., & Umberson, D. (1988, July 29). Social helper, and cytolytic lymphocytes. Biochemical Pharmacology, 35, relationships and health. Science, 241, 540-545. 1137-1142. lronson, G., Friedman, A., Klimas, N., Antoni, M., Fletcher, M. A., Kiecolt-Glaser, ,l. K., Dura, ,l. R., Speicher, C. E., Trask, O. J., & Glaser, LaPerriere, A., Simoneau, J., & Schneiderman, N. (1994). Distress, R. G. ( 1991 ). Spousal caregivers of dementia victims: Longitudinal denial, and low adherence to behavioral interventions predict faster changes in immunity and health. Psychosomatic Medicine, 53, 345- disease progression in gay men infected with human immunodefl- 362. ciency virus. International Journal of Behavioral Medicine, 1, 90- Kiecolt-Glaser, J. K., & Glaser, R. (1988a). Immunological compe- 105. tence. In E. A. Blechman & K. D. Brownwell (Eds.), Handbook of Irwin, M., Brown, M, Patterson, T., Hauger, R., Mascovich, A., & behavioral rnedicinefor women (pp. 195-205 ). New York: Pergamon Grant, I. ( 1991 ). Neuropeptide Y and natural killer cell activity: Press. Findings in depression and Alzheimer caregiving stress. FASEB Jour- Kiecolt-Glaser, J. K, & Glaser, R. (1988b). Methodological issues in nal 5, 3100-3107. the behavioral immunology research with humans. Brain. Behavior, Jackson, L. A., & Adams-Campbell, L. L. (1994). John Henryism and and lmmunity, 2, 67-78. blood pressure in Black college students. Journal of Behavioral Med- Kiecolt-Glaser, J. K., & Glaser, R. (1989). Interpersonal relationships icine, 17, 69-79. and immune function. In L, Carstensen & J. Neale (Eds.), Mecha- Jait~, B. M., & Behrman, H. R. (1979). Methods of hormone radioim- nisms ofpsychological influence on physical health (pp. 43-59 ). New munoassay, New York: Academic Press. York: Plenum Press. James, S. A., LaCroix, A. Z., Kleinbaum, D. G., & Strogatz, D. S. Kiecolt-Glaser, J. K., & Glaser, R. (1995). Psychoneuroimmunology (1984). John Henryism and blood pressure differences among Black and health consequences: Data and shared mechanisms. Psychoso- men: ti. The role of occupational stressors. Journal of Behavioral matic Medicine, 57, 269-274. Medicine, 7, 259-275. Kiecolt-Glaser, J. K., Glaser, R., Williger, D., Stout, J., Messick, G., .lanes, C. R. (1990). Migration, changing gender roles and stress: The Sheppard, S, Ricker, D., Romisher, S. C., Briner, W., Bonnell, G., & Samoan case. Medical Anthropotog); 12, 217-248. Donnerberg, R. (1985). Psychosocial enhancement ofimmunocom- Janes, C. R., & Pawson, I. G. (1986). Migration and biocultural adap- petence in a geriatric population. Health Psychology, 4. 25-41. tation: Samoans in California. Social Science and Medicine, 22, 821- Kiecolt-Gtaser, J. K., & Greenberg, B. (1984). Social support as a mod- 834. erator of the aftereffects of stress in female psychiatric inpatients. Jemmott, J. B., Ili, & Magloire, K. ( 1988 ). Academic stress, social sup- Journal of Personality and Social Psychology, 93, 192-199. port, and secretory immunogiobin A. Journal of Personality and So- Kiecolt-Glaser, J. K., Malarkey, W. B., Chee, M., Newton, T., Cacioppo, cial Psychology; 55, 803-810. J. T., Map, H., & Glaser, R. ( 1993 ). Negative behavior during marital Jung, J. (1990). Global versus health-specific social support and match of preferred and perceived social support levels in relationship to conflict is associated with immunological down-regulation. Psycho- compliance and blood pressure in hypertensives. Journal of Applied somatic Medicine, 55, 395-409. Social Psychology, 20, 1103-1111. Kirschbaum, C., Klauer, T, Filipp, S. H., & Hellhammer, D. H. (1995). Kamarck, T. W. (1992). Recent developments in the study of cardio- Sex-specific effects of social support on cortisol and subjective re- vascular reactivity: Contributions from psychometric theory and so- sponses to acute psychological stress. Psychosomatic Medicine, 57, cial psychology. Psychophysiology, 29, 491-503. 23-31. Kamarck, T. W., Annunziato, B., & Amateau, L. M. ( 1995 ). Affiliation Knox, S. S. (1993). Perception of social support and blood pressure in moderates the effects of social threat on stress-related cardiovascular young men. Perceptual andMotor Skills, 77, 132-134. responses: Boundary conditions for a laboratory model of social sup- Knox, S. S., Theorell, T., Svensson, J. C., & Wailer, D. (1985). The port. Psychosomatic Medicine, 57, 183-194. relation of social support and working environment to medical vari- Kamarck, T. W., & Jennings, R. J. (1991). Biobehavioral factors in ables associated with elevated blood pressure in young males: A sudden cardiac death. Psychological Bulletin, 109, 42-75. structural model. Social Science and Medicine, 21, 525-531. Kamarck, T. W., Jennings, J. R., Debski, T. T., Glickman-Weiss, E., Krantz, D. S., Helmers, K. E, Bairey, N., Nebel, L. E., Hedges, S. M., Johnson, P. S., Eddy, M. J., & Manuck, S. B. (1992). Reliable mea- & Rozanski, A. ( 1991 ). Cardiovascular reactivity and mental stress- sures of behaviorally-evoked cardiovascular reactivity from a PC- induced myocardial ischemia in patients with coronary artery dis- based test battery: Results from student and community samples. ease. Psychosomatic Medicine, 53, 1-12. Psychophysiology, 29, 17-28. Krantz, D. S., & Manuck, S. B. (1984). Acute physiologic reactivity and Kamarck, T. W., Manuck, S. B., & Jennings, J. R. (1990). Social sup- risk of cardiovascular disease: A review and methodologic critique. port reduces cardiovascular reactivity to psychological challenge: A Psychological Bulletin, 96, 435-464. laboratory model. Psychosomatic Medicine, 52, 42-58. Larsen, P. B., Schneiderman, N., & Pasin, R. D. (1986). Physiological Kart, C. S., Metress, E. K., & Metress, S. P. (1992). Human aging and bases of cardiovascular psychophysiology. In M. G. H. Coles, E. Don- chronic disease. Boston: Jones & Bartlett. chin, & S. W. Porges (Eds.), Psychophysiology: System& processes. Kasl, S. V., & Cobb, S. (1980). The experience of losing a job: Some and applications (pp. 122-165 ). New York: Guilford Press. SOCIAL SUPPORT AND PHYSIOLOGY 529

Lehrman, D. (1964). The reproductive behavior of ring doves. Scien- individual differences. In N. Schneiderman, S. B. Weiss, & P. Kauf- tific American, 211, 48-54. mann (Eds.), Handbook of methods anat measurements in cardiovas- Lepore, S. J. (1992). Social conflict, social support, and psychological cular behavioral medicine (pp. 365-382). New York: Plenum Press. distress: Evidence of cross-domain buffering effects. Journal of Per- Marsland, A. L, Manuck, S. B., Fazzari, T. V., Stewart, C. J., & Rabin, sonality and Social Psychology, 63, 857-867. B. S. (t995). Stability of individual differences in cellular immune Lepore, S. J. ( 1995 ). Cynicism, social support, and cardiovascular re- responses to acute psychological stress. Psychosomatic Medicine, 57, activity. Health Psychology, 14, 210-216. 295-298. Lepore, S. J., Allen, K. A., & Evans, G. W. (1993). Social support low- Martinez-Vargas, M. C., & Erikson, C. J. (1973). Some social and hor- ers cardiovascular reactivity to an acute stressor. Psychosomatic Med- monal determinants of nest-building behaviour in the ring dove icine, 55, 518-524. ( Streptopelia risoria ). Behaviour, 45, 12-37. Lercher, P., Hormagl, J., & Kofler, W. W. (1993). Work noise annoy- Matthews, K. A., Weiss, S. M., Detre, T., Dembroski, T. M., Falkner, ance and blood pressure: Combined effects with stressful working B., Manuck, S. B., & Williams, R. B. (1986). Handbook of stress, conditions. International Archives of Occupational and Environmen- reactivity, and cardiovascular disease. New York: Wiley. tal Health, 65, 23-28. McGonigle, M. M., Smith, T. W., Benjamin, L. S., & Turner, C. W. Levine, D. M., Green, L. W., Deeds, S. G., Chwalow, J., Russell, (1993). Hostility and nonshared family environment: A study of R. P., & Finlay, J. (1979). Health education for hypertensive patients. monozygotic twins. Journal of Research in Personality, 27, 23-34. Journal of the American Medical Association. 241, 1700-1703. Mclntosh, W. A., Kaplan, H. B., Kubena, K. S., & Landmann, W. A. Levy, S. M., Herberman, R. B., Whiteside, T., Sanzo, K., Lee, J., & ( 1993 ). Life events, social support, and immune response in elderly Kirkwood, J. (1990). Perceived social support and tumor estrogen/ individuals. International Journal ofAging and Human Development, progesterone receptor status as predictors of natural killer cell activ- 37, 23-36. ity in breast cancer patients. Psychosomatic Medicine, 52, 73-85. McKinnon, W., Baum, A., & Morokoff, P. (1988). Neuroendocrine Lewis, J. W., Shavit, Y., Terman, G. W., Nelson, L. R., Gale, R. P., & measures of stress. In H. L, Wagner (Ed.), Social psychophysiology Liebeskind, J. C. (1983). Apparent involvement of opioid peptides and emotion (pp. 43-64). New York: Wiley. in stress-induced enhancement of tumor growth. Peptides, 4, 635- McNaughton, M. E., Smith, L. W., Patterson, T. L., &Grant, I. (1990). 638. Stress, social support, coping resources, and immune status in elderly Light, K. C. ( 1981 ). Cardiovascular responses to effortful active coping: women. Journal of Nervous and Mental Disease, 178, 460-461. Implications for the role of stress in hypertension development. Psy- Medline [Database]. (1983-1995, June 1 ). Bethesda, MD: National chophysiolog~; 18, 216-225. Library of Medicine [ Producer]. Available from: SilverPlatter. Light, K. C., Dotan, C. A., Davis, M. R., & Sherwood, A. (1992). Car- Meites, J., Goya, R., & Takahashi, S. ( 1987 ). Why the neuroendocrine diovascular responses to an active coping challenge as predictors of system is important in aging processes. Experimental Gerontology, blood pressure patterns 10 to 15 years later. Psychosomatic Medicine, 22, 1-15. 54, 217-230. Mills, P. J., & Dimsdale, J. J. (1993). The promise of adrenergic recep- Linden, W., Chambers, L., Maurice, J., & Lenz, J. W. (1993). Sex tor studies in psychophysiologic research. II: Applications, limita- differences in social support, self-deception, hostility, and ambula- tions, and progress. Psychosomatic Medicine, 55, 448-457. tory cardiovascular activity. Health Psychology, 12, 376-380. Morisk2~; D. E., DeMuth, N. M., Field-Fass, M., Green, L. W., & Lev- Linn, B. S:, Linn, M. W., & Klimas, N. G. (1988). Effects of psycho- ine, D. M. (1985). Evaluation of family health education to build physical stress on surgical outcomes. Psychosomatic Medicine, 50, social support for long-term control of high blood pressure. Health 230-244. Education Quarterly, 12, 35-50. Livingston, I. L., Levine, D. M., & Moore, R. D. (1991). Social inte- Morisky, D. E., Levine, D. M., Green, L. W., Shapiro, S., Russell, P., & gration and Black intraracial variations in blood pressure. Ethnicity Smith, C. R. ( 1983 ). Five-year blood pressure control and mortality andDisease, 1, 135-149. following health education for hypertensive patients. American Jour- Lumio, J., Welin, M.-G., Hirvonen, P., & Weber, T. (1983). Lympho- nal of Public Health, 73, 153-162. cyte subpopulations and reactivity during and after infectious mono- Mullin, B. (1989). Advanced BASIC meta-analysis. Hillsdale, NJ: nucleosis. Medical Biology, 61, 208-213. Erlbaum. Madden, K. S, & Livnat, S. ( 1991 ). Catecholamine action and immu- Murasko, D. M., Weiner, P., & Kaye, D. (1988). Association of lack of nologic reactivity. In R. Ader, D. L. Felten, & N. Cohen (Eds.), Psy- mitogen-induced lymphocyte proliferation with increased mortality choneuroimmunology (pp. 283-310). New York: Academic Press. in the elderly. Aging: Immunology and Infectious Disease, 1, 1-6. Malarkey, W. B., Kiecolt-Glaser, J. K., Pearl, D., & Glaser, R. (1994). Orth-Gomer, K., Rosengren, A., & Wilhelmsen, L. ( 1993 ). Lack of so- Hostile behavior during marital conflict affects pituitary and adrenal cial support and incidence of coronary heart disease in middle-aged hormones. Ps)~chosomaticMedicine, 56, 41-51. Swedish men. Psychosomatic Medicine, 55, 37--43. Malcolm, A. T., & Janisse, M. P. ( 1991 ). Additional evidence for the Orth-Gomer, K., & Unden, A. ( 1987 ). The measurement of social sup- relationship between Type A behavior and social support in men. Be- port in population surveys. Social Science and Medicine, 24, 83-94. havioral Medicine, 17, 131- t 34. Pagel, M. D., Erdly, W. V~, & Becker, J. (1987). Social networks: We get Manuck, S. B. (1994). Cardiovascular reactivity in cardiovascular dis- by with (and in spite of) a little help from our friends. Journal of ease: "Once more unto the breach." International Journal of Behav- Personality and Social Psychology, 53, 793-804. ioral Medicine, 1, 4-3 I. Paxton, H., Kidd, P., Landay, A., Giorgi, J., Flomenberg, N., Walker, E., Manuck, S. B., Cohen, S. C., Rabin, B., Muldoon, M. E, & Bachen, Valentine, E, Fahey, J., & Gelman, R. (1989). Results of the flow E. A. ( 1991 ). Individual differences in cellular immune response to cytometry ACTG quality control program: Analysis and findings. stress. Psychological Science, 2, 111-115. Clinical lmmundogy and lmmunopathology, 52, 68-84. Manuck, S. B., Kaplan, J. R., & Clarkson, T. B. (1983). Behaviorally Perry, S., Fishman, B., Jacobsberg, L., & Frances, A. (1992). Relation- induced heart rate reactivity and atherosclerosis in Cynomolgus ships over 1 year between lymphocyte subsets and psychosocial vari- monkeys. Psychosomatic Medicine, 45, 95-108. ables among adults with infection by human immunodeficiency vi- Manuck, S. B., Kasprowicz, A. L., Monroe, S. M., Larkin, K. T., & rus. Archives of General Psychiatry, 49, 396-401. Kaplan, J. R. (1989). Psychophysiologic reactivity as a dimension of Persson, L., Gullberg, B., Hanson, B. S., Moestrup, T., & Ostergren, 530 UCHINO, CACIOPPO, AND KIECOLT-GLASER

E O. (1994). HIV infection: Social network, social support, and CD4 Schleifer, S. J., Keller, S. E., Bond, R. N., Cohen, J., & Stein, M. (1989). lymphocyte values in infected homosexual men in Malmo, Sweden. Major depressive disorder and immunity: Role of age, sex, severity, Journal of Epidemiology and Community Health, 48, 580-585. and hospitalization. Archives of General Psychiatry, 46, 81-87. Pierce, G. R., Sarason, I. G., & Sarason, B. R. (1991). General and Schlesinger, M., & Yodfat, T. ( 1991 ). The impact of stressful life events relationship-based perceptions of social support: Are two constructs on natural killer cells. Stress Medicine, 7, 53-60. better than one? Journal of Personality and Social Psychology, 61, Seeman, T. E., Berkman, L. F., Blazer, D., & Rowe, J. W. (1994). Social 1028-1039. ties and support and neuroendocrine function: The MacArthur stud- Pinto, R. P., Sirota, A. D., & Brown, W. A. (1985). Behavioral instruc- ies of successful aging. Annals ofBehavioral Medicine, 16, 95-106. tion for a patient's relative to increase antihypertensive compliance. Seeman, T. E., & Robbins, R. J. (1994). Aging and the hypothalamic- Psychosomatics, 26, 677-679. pituitary-adrenal response to challenge in humans. Endocrine Re- Plaut, M. (1987). Lymphocyte hormone receptors. Annual Review of views, 15, 233-260. Immunology, 5, 621-669. Sgoutas-Emch, S. A., Cacioppo, J. T., Uchino, B. N., Malarkey, W., Porges, S. W. (1992). Vagal tone: A physiological marker of stress vul- Pearl, D., Kiecolt-Glaser, J. K., & Glaser, R. (1994). The effects of an nerability. Pediatrics, 90, 498-504. acute psychological stressor on cardiovascular, endocrine, and cellu- PsycLIT [ Database ]. ( 1974-1995, June). Washington, DC: American lar immune response: A prospective study of individuals high and low Psychological Association [Producer]. Available from: SilverPlatter. in heart rate reactivity. Psychophysiology, 31, 264-271. Quittner, A. L., Glueckauf, R. L., & Jackson, D. N. (1990). Chronic Sheffield, D., & Carroll, D. (1994). Social support and cardiovascular parenting stress: Moderating versus mediating effects of social sup- reactions to active laboratory stressors. Psychology and Health, 9, port. Journal of Personality and Social Psychology,, 59, 1266-1278. 305-316. Roberts-Thomson, I. C., Whittingham, S., Youngchaiyud, U., & Shekelle, R. B., Gale, M., Ostfeld, A. M., & Paul, O. (1983). Hostility, Mackay, I. R. (1974). Aging, immune response, and mortality. Lan- risk of coronary heart disease, and mortality. Psychosomatic Medi- cet, ii, 368-370. cine, 45, 109-114. Rook, K. S. (1984). The negative side of social interaction: Impact on Sheridan, J. E, Feng, N., Bonneau, R., Malarkey, W., & Hughes, J. psychological well being. Journal of Personality and Social Psychol- (1991). Restraint stress differentially affects anti-viral cellular and ogy, 46, 1097-1108. humoral immune responses in mice. Journal of Neuroimmunology, Rook, K. S., & Pietromonaco, P. (1987). Close relationships: Ties that 31,245-255. heal or ties that bind. Advances in PersonalRelationships, 1, 1-35. Sherwood, A., Allen, M. T., Fahrenberg, J., Kelsey, R. M., Lovallo, Rose, R. N., Gordon, T. P., Bernstein, I. S. (1972, November 10). W. R., & Doornen, L. J. P. van. (1990). Methodological guidelines Plasma testosterone levels in the male rhesus: Influences of sexual for impedance cardiography. Psychophysiology, 27, 1-23. and social stimuli. Science, 178, 643-645. Sherwood, A., Dolan, C. A., & Light, K. C. (1990). Hemodynamics of Rosenthal, R. (1984). Meta-analytic procedures of social research. blood pressure responses during active and passive coping. Psycho- London: Sage. physiology, 27, 656-668. Rozanski, A., Bairey, N., Krantz, D. S., Friedman, J., Resser, K. J., Mo- Shumaker, S. A., & Hill, D. R. (1991). Gender differences in social rell, M., Hilton-Chalfen, S., Herstrin, L., Bietendorf, J., & Berman, support and health. Health Psychology, 10, 102-111. D. S. (1988). Mental stress and the induction of silent myocardial Sibinga, N. E. S., & Goldstein, A. (1988). Opioid peptides and opioid ischemia in patients with coronary artery disease. New England Jour- receptors in cells of the immune system. Annual Review oflmmunol- nal ofMedicine, 318, 1005-1012. ogy, 6, 219-249. Ruehlman, L. S., & Karoly, P. (1991). With a little flak from my Smith, C. E., Fernengel, K., Holcroft, C., Gerald, K., & Marien, L. friends: Development and preliminary validation of the Test of Neg- (1994). Meta-analysis of the association between social support and ative Social Exchange (TENSE). Psychological Assessment, 3, 97- health outcomes. Annals ofBehavioral Medicine, 16, 352-362. 104. Smith, E. L. (1984). Special considerations in developing exercise pro- Rupprecht, R., Wodarz, N., Kornhuber, J., Schmitz, B., Wild, K., grams for the older adult. In J. D. Matarazzo, S. M. Weiss, J. A. Herd, Braner, H. U., Muller, O. A., & Riederer, P. (1990). In vivo and in N. E. Miller, & S. M. Weiss (Eds.), Behavioral health:A handbook of vitro effects of glucocorticoids on lymphocyte proliferation in man: health enhancement and disease prevention (pp. 525-546). New Relationship to glucocorticoid receptors. Neuropsychobiology, 24, York: Wiley. 61-66. Smith, J. J., & Kampine, J. P. (1990). Circulatoryphysiology London: Russell, D. W., & Cutrona, C. E. (1991). Social support, stress, and Williams & Wilkins. depressive symptoms among the elderly: Test of a process model. Psy- Smith, T. W. (in press). Assessment and modification of coronary prone chology and Aging, 6, 190-201. behavior: A transactional view of the person in social context. In Sallis, J. E, Trevorrow, T. R., Johnson, C. C., HoveU, M. E, & Kaplan, A. J. Goreczny (Ed.), Handbook of health and rehabilitation psychol- R. M. (1987). Worksite stress management: A comparison of pro- ogy New York: Plenum Press. grams. Psychology and Health, 1, 237-255. Smith, T. W., & Frohm, K. D. (1985). What's so unhealthy about hos- Sanders, V. M., & Munson, A. E. ( 1985 ). Norepinephrine and the anti- tility? Construct validity and psychosocial correlates of the Cook and body response. PharmacologicalReviews, 37. 229-248. Medley Ho Scale. Health Psychology, 4, 503-520. Sarason, B. R., Sarason, I. G., & Pierce, G. R. (1989). Social support: Smith, T. W., Pope, M. K., Sanders, J. D., Allred, K. D., & O'Keeffe, An interactional view. New York: Wiley. J. L. ( 1988 ). Cynical hostility at home and work: Psychosocial vul- Sarason, I. G., Sarason, B. R., & Shearin, E. N. (1986). Social support nerability across domains. Journal of Research in Personality, 22, as an individual difference variable: Its stability, origins, and rela- 525-548. tional aspects. Journal ofPersonality and Social Psychology, 50, 845- Snyder, B. K., Roghmann, K. J., & Sigal, L. H. (1990). Effects of stress 855. and other biopsychosocial factors on primary antibody response. Schenker, E. L., Hultin, L. E., Bauer, K. D., Ferbas, J., Margolick, J. B., Journal ofAdolescent Health Care, 11,472-479. & Giorgi, J. V. (1993). Evaluation of a dual-color flow cytometry Snyder, B. K., Roghmann, K. J., & Sigal, L. H. (1993). Stress and psy- immunophenotyping panel in a multicenter quality assurance pro- chosocial factors: Effects on primary cellular immune response. gram. Cytometry, 14, 307-317. Journal of Behavioral Medicine, 16, 143-161. SOCIAL SUPPORT AND PHYSIOLOGY 531

Snydersmith, M. A., & Cacioppo, J. T. (1992). Parsing complex social killer cell activity during cortisol and adrenaline infusion in healthy factors to determine component effects: I. Autonomic activity and volunteers. European Journal of Clinical Investigation, 17, 497-503. reactivity as a function of human association. Journal of Social and Uchino, B. N., Cacioppo, J. T., Malarkey, W., & Glaser, R. (1995). In- Clinical Psychology, 11, 263-278. dividual differences in cardiac sympathetic control predict endocrine Solomon, Z., Mikulincer, M., & Hobfoll, S. E. (1986). Effects of social and immune responses to acute psychological stress. Journal of Per- support and battle intensity on loneliness and breakdown during sonality and Social Psychology, 69, 736-743. combat. Journal of Personality and Social Psychology, 51, 1269- Uchino, B. N., Cacioppo, J. T., Malarkey, W. M., Glaser, R., & Kiecolt- 1276. Glaser, J. K. (1995). Appraisal support predicts age-related differ- Spitzer, S. B., Llabre, M. M., Ironson, G. H., Gellman, M. D., & Schnei- ences in cardiovascular function in women. Health Psychology, 14, derman, N. ( 1992 ). The influence of soeial situations on ambulatory 556-562. blood pressure. Psychosomatic Medicine, 54, 79-86. Uchino, B. N., Kiecolt-Glaser, J. K., & Cacioppo, J. T. (1992). Age- Stahl, S. M., Kelley, C. R., Neill, P., Grim, C. E., & Mamlin, J. (1984). related changes in cardiovascular response as a function of a chronic Effects of home blood pressure measurement on long-term BP con- stressor and social support. Journal of Personality and Social Psy- trol. American Journal of Public Health, 74, 704-709. chology, 63, 839-846. Stavig, G. R., lgra, A., & Leonard, A. R. (1984). Hypertension among Uchino, B. N., Kiecolt-Glaser, J. K., & Cacioppo, J. T. (1994). Con- Asians and Pacific Islanders in California. American Journal of Epi- struals of preillness relationship quality predict cardiovascular re- demiology, 119, 677-691. sponse in family caregivers of Alzheimer's disease victims. Psychology Stokes, J. P. (1985). The relation of social network and individual andAging, 9, 113-120. differences in loneliness. Journal of Personality and Social Psychol- Umberson, D. ( 1987 ). Family status and health behaviors: Social con- ogy, 48, 981-990. trol as a dimension of social integration. Journal ofHealth and Social Stone, A. A., Cox, D. S., Vaidimarsdottir, H., & Neale, J. M. (1987). Behavior, 28, 306-319. Secretory lgA as a measure of immunocompetence. Journal of Hu- Unden, A.-L., Orth-Gomer, K., & Elofsson, S. ( 1991 ). Cardiovascular man Stress, 13, 136-140. effects of social support in the work place: Twenty-four hour ECG Strogatz, D. S., & James, S. A. ( 1986 ). Social support and hypertension monitoring of men and women. Psychosomatic Medicine, 53, 50-60. among Blacks and Whites in a rural, southern community. American Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and val- Journal of Epidemiology, 124, 949-956. idation of brief measures of positive and negative affect: The PANAS Szklo, M. (1979). Epidemiologic patterns of blood pressure in children. scales. Journal of Personality and Social Psychology, 54, 1063-1070. Epidemiologic Reviews, 1, 143-169. Watson, D., & Pennebaker, J. W. (1989). Health complaints, stress, and Tardy, C. H. ( 1985 ). Social support measurement. American Journal distress: Exploring the central role of negative affectivity. Psychologi- of Community Psychology, 13, 187-202. caI Review, 96, 234-254. Tardy, C. H., Thompson, W. R., & Allen, M. T. (1989). Cardiovascular Whiteside, T. L., & Herberman, R. B. (1994). Role of human natural i'esponses during speech: Does social support mediate the effects of killer cells in health and disease. Clinical and Diagnostic Laboratory talking on blood pressure? Journal of Language and Social Psycho# Immunology, 1, 125-133. ogy, 8, 271-285. Williams, L. T., Snyderman, R., & Lefkowitz, R. J. (1976). Identifica- Theorell, T., Blomkvist, V., Jonsson, H., Schulman, S., Berntorp, E., & tion of b-adrenergic receptors in human lymphocytes by (-) [-3H ] Stigendel, L. ( 1995 ). Social support and the development of immune alprenolol binding. Journal of Clinical Investigation, 5 7, 149-155. function in human immunodeficiency virus infection. Psychoso- Winemiller, D. R., Mitchell, E. M., Sutliff, J., & Cline, D. J. (1993). matic Medicine, 57, 32-36. Measurement strategies in social support: A descriptive review of the Theorell, T., Orth-Gomer, K., & Eneroth, P. (1990). Slow-reacting im- literature. Journal of Clinical Psychology, 49, 638-648. munoglobin in relation to social support and changes in job strain: A Winnubst, J. A. M., Marcelissen, E H. G., & Kleber, R. J. (1982). preliminary note. Psychosomatic Medicine, 52, 511-516. Effects of social support in the stressor-strain relationship: A Dutch Thomas, P. D., Goodwin, J. M., & Goodwin, J. S. (1985). Effects of sample. Social Science and Medicine, 16, 475-482. social support on stress-related changes in cholesterol level, uric acid level, and immune function in an elderly sample. American Journal Received February 14, 1995 of Psychiatry, 142, 735-737. Revision received July 21, 1995 Tonnesen, E., Christensen, N. J., & Brinklov, M. M. (1987). Natural Accepted July 24, 1995 •