social support in stressful and non-stressful work units

17
SocialBehaviour,Vol. 4, 155-171 (1989) Stress and Afïïliation Reconsidered:the Effects of SocialSupport in Stressful and Non-stressful Work Units BRAM P. BUUNK, PETER P.M. JANSSEN and NICO W. VANYPEREN Department of Psychology, University of Nijmegen, Postbox 9104, ó500 HE Nijmegen, The Netherlands ABSTRACT The literature concerning the role of social support in reducing stress at work is critically reviewed. It is concluded that, in many cases, social support does not seem to have an effect, or sometimes even a negative effect, upon stress reactions. Explanations for these inconsistent findings are offered on the basis of various social psychological theories. In addition, it is pointed out that a major drawback of most occupational stressstudies is that the assessment of stressors,social support and strains are all based upon the perception of the respondent. In the present study, eight stressors were included, and for each stressor, the four departments that were characterized by the highest mean score on a certain stressor (ftlglr stressorunits) were selected, as well as the four departments that had the lowest score (low stressor unils). According to the buffering hypothesis, support from one's supervisor and colleaguesshould have a stronger relationship with strains within the high stressorunits than in the low stressor units. It was found that support served particularly as a buffer against the negative consequences of inability to leave the workplace, role overload, lack of participation and job future uncertainty, and that support seemed mainly to reduce cognitive anxiety and negative affect, but was barely related to other strains. Remarkably, some evidence was also found for negative effects of social support. The results are discussed in the context of the literature on stress and affiliation and on social influence. Key words: Social support, occupational stress, affiliation, social influence, social comparison, helping. INTRODUCTION The classical experiments on affiliation under stress conducted by Schachter(1959) nearly thirty years ago, have found their contemporary counterpart in field research examining the effects of social support upon health and well-being. Over the past decades, numerous studies have explored the link between features of social relations and networks on the one hand, and a variety of physical and psychological symptoms and the other, particularly among individuals who are confronted with important life events such as divorce, pregnancy, chronical diseaseand unemploy- Note: The authoÍs thank Hedy Kleijweg for her editorial assistance. 08854249 I89 /03015 s-r7 $08. 50 O 1989 by John Wiley & Sons, Ltd. Receíved 19lanuary 1989 Accepted25 May 1989

Upload: others

Post on 09-Feb-2022

12 views

Category:

Documents


0 download

TRANSCRIPT

Social Behaviour, Vol. 4, 155-171 (1989)

Stress and Afïïliation Reconsidered: the Effects ofSocial Support in Stressful and Non-stressful WorkUnits

BRAM P. BUUNK, PETER P.M. JANSSEN and NICO W. VANYPERENDepartment of Psychology, University of Nijmegen, Postbox 9104, ó500 HE Nijmegen, TheNetherlands

ABSTRACT

The literature concerning the role of social support in reducing stress at work is criticallyreviewed. It is concluded that, in many cases, social support does not seem to have an effect,or sometimes even a negative effect, upon stress reactions. Explanations for these inconsistentfindings are offered on the basis of various social psychological theories. In addition, it ispointed out that a major drawback of most occupational stress studies is that the assessmentof stressors, social support and strains are all based upon the perception of the respondent.In the present study, eight stressors were included, and for each stressor, the fourdepartments that were characterized by the highest mean score on a certain stressor (ftlglrstressor units) were selected, as well as the four departments that had the lowest score (lowstressor unils). According to the buffering hypothesis, support from one's supervisor andcolleagues should have a stronger relationship with strains within the high stressor units thanin the low stressor units. It was found that support served particularly as a buffer against thenegative consequences of inability to leave the workplace, role overload, lack ofparticipation and job future uncertainty, and that support seemed mainly to reduce cognitiveanxiety and negative affect, but was barely related to other strains. Remarkably, someevidence was also found for negative effects of social support. The results are discussed inthe context of the literature on stress and affiliation and on social influence.

Key words: Social support, occupational stress, affiliation, social influence, socialcomparison, helping.

INTRODUCTION

The classical experiments on affiliation under stress conducted by Schachter (1959)nearly thirty years ago, have found their contemporary counterpart in fieldresearch examining the effects of social support upon health and well-being. Overthe past decades, numerous studies have explored the link between features of socialrelations and networks on the one hand, and a variety of physical and psychologicalsymptoms and the other, particularly among individuals who are confronted withimportant life events such as divorce, pregnancy, chronical disease and unemploy-

Note: The authoÍs thank Hedy Kleijweg for her editorial assistance.

08854249 I 89 / 03015 s-r7 $08. 50O 1989 by John Wiley & Sons, Ltd.

Receíved 19 lanuary 1989Accepted25 May 1989

156 B.P. Buunk et al.

ment (for recent reviews see Barrera, 1986; Cohen and Wills, 1985; Kessler, Priceand Wortman, 1985; Sarason and Sarason, 1985).

At the onset of the eighties researchers investigating occupational stress alsobecame interested in this subject. The theoretical analyses and research conductedby House (1981) have served as an important impetus in this area. To date, morethan 100 studies have examined the stress-alleviating role of social support at work.Nevertheless, as will become apparent in this paper, the results of these researchefforts are quite equivocal. Many contradictory findings have been reported, and inmany cases no effects of social support have been found. Furthermore, mostresearch in this area has been a-theoretical and has focused little attention upon thesocial psychological processes that might be relevant.

OCCUPATIONAL STRESS

Contrary to most of the stress research within psychology, the study of occupationalstress has its origin within social psychology, and dates back to the fifties, whenthe first major and still very influential program on occupational stress was initiatedby Robert Kahn and his colleagues at the University of Michigan. In this programtheimportanceof the subjective socialenvironmentwasemphasized,i.e. thewayinwhich the organization is perceived by the individual (French and Kahn 1962;Kahn, Wolfe, Quinn, Snoek and Rosenthal, 1964). This concept was directlyderived from the Lewinian notion of psychological life space, which is how Lewinreferred to the individual's experience of the environment that was considered tohave a direct impact upon behavior. In addition to Lewinian thinking, role theoryfigured prominently in the pioneering research on occupational stress, and has untilnow prompted the majority of occupational stress studies to focus upon role-related'stressors', including role conflict, role ambiguity and role overload (for a reviewsee Sell, Brief and Schuler, 1981).

Although the concept 'social support' was not yet mentioned as such in the earlyresearch on occupational stress, its relevance was recognized from the beginning.For instance, Zander and Quinn (1962) reviewed a series of field studies andexperiments in the tradition of group dynamics that pointed to the importance ofinterpersonal attraction between workers, as well as clear and considerateleadership in the prevention of feelings of nervousness among workers. In theirpioneer study, Kahn el al. (196$ highlighted the importance of social support bydocumenting the fact that features of interpersonal relations at work, such ascommunication, dependence and power, moderated the negative emotionalconsequences of role conflict and ambiguity.

The conceptual model developed at the University of Michigan in the beginningof the sixties, still remains, in a moderated way, an important framework forresearch on occupational stress. This model assumes that certain stressors, such asrole ambiguity, role conflict and future uncertainty, may occur in the subjectiveenvironment of the individual. It is assumed that if these stressors are more or lessof a permanent nature and overburden the coping potential of the individual, alarge array of psychological, physiological and behavioral stress reactions (e.g.worrying, headaches, increased blood pressure, raised cholesterol level, excessivesmoking, and psychological or psycho-somatic complaints) may result (see Van

Stress and affiliation reconsidered 157

Dijkhuizen, 1980; Kahn 1981). It is further hypothesized that the emergence ofthese reactions depends on personality traits such as Type A behavior, neuroticism,and rigidity, and on features of the social environment of individuals, particularlysocial support from one's boss, colleagues, partner and friends.

SOCIAL SUPPORT: PERSPECTIVES AND EFFECTS

With regard to the effects of social support, a distinction is generally made betweenbuffering effects and direct effects (c/. Cohen and Wills, 1985). Direcl effects are, instatistical terms, main effects of social support upon well-being and health. Sucheffects occur when individuals who are involved in supportive and satisfyingrelationships have a relatively high degree of health and well-being. Social supportresearchers have, however, been particularly interested in buffer effects, and haveassumed that support can counteract the negative consequences of stress uponhealth and well-being. A buffer effect is, in statistical terms, a specific interactioneffect: it refers to those instances where a particularly strong relationship betweensocial support and health is found among individuals who are confronted withstressful conditions. Research on both types of effects within organizations will betreated in this paper.

Despite the relevance attributed to social support in research in occupationalstess, there is limited agreement as to the exact conceptualization and definition ofthis concept, and several perspectives on this issue can be distinguished (Winnubst,Buunk and Marcelissen, 1988). First, we can observe the perspective that arguesthat adequate social support is equal to the quality of interpersonal relationships orthe organizational climate. Others, however, (e.g. Rook, 1984) have argued thatthe concept ofsocial support should be confined to helping behaviors. Indeed, froma second perspective, social support is defined in terms of specific helpingbehaviors, i.e. the flow of certain elements between individuals, such as emotionalconcern, feedback, information and instrumental aid (House, 1981). It must beemphasized, however, that when social support is assessed in survey research as theamount of help an individual has received, positive correlations between supporrand distress might be expected, not because help has a negative effect but simplybecause the individual under stress more often seeks help. Indeed, such positivecorrelations have been reported in studies using such measures (Barrera, 1986;Cohen and Wills, 1985). For the purpose of this study, we employed a thirdperspective, namely that social support is the degree to which others are perceivedas helpful and dependable. From this perspective, social support constitutes theappraisal that, in case of stressful circumstances, others can be relied upon foradvice, information and empathic understanding, guidance and alliance (Sarasonand Sarason, 1985). In many studies on occupational stress, measures addressingsuch variables have been used (Winnubst, Buunk and Marcelissen, 1988).

Perceived helpfulness may buffer stress through a process that can be designatedas social inoculation (Buunk, 1989). This concept refers to the notion that someonewho perceives the opportunity. for help from others has a resource at his or herdisposal that makes him or her relatively immune to stress, for instance by helpinghim or her to redefine a situation as less threatening. Buffering effects of supportwith regard to stressful life events seem to occur particularly when the perceived

158 B.P. Buunk et al.

availability of interpersonal resources is employed as an indicator of social support(Cohen and Wills, 1985). There is also some experimental evidence for the notionof social inoculation. For instance, Sarason and Sarason (1986) found that subjectswho were informed they could turn to the experimenter for help (but actually neverdid) were better able to perform on a task that required considerable cognitiveeffort than subjects who did not have this opportunity.

DIRECT EFFECTS OF SOCIAL SUPPORT AT WORK

Is there clear evidence that rewarding relationships with superiors and colleaguesare related to the absence of stress? Are there associations between measures ofsocial support at work and indicators of mental and physical well-being? A firstremarkable finding is that in many studies that include relevant measures, nocorrelations between support and indexes of mental or physical health arepresented, sometimes because the research was primarily aimed at assessing buffereffects (e.g. Burke, 1982; Brenner, Sorbom and Wallins, 1985; LaRocco, Houseand French, 1980; Payne and Fletcher, 1983). In many other studies, moderatenegative correlations have been established between social support and psycho-logical stress reactions. For instance, in one of the better designed studies, Caplan,Cobb, French, van Harrison and Pinneau (1975) found that a lack of support fromone's supervisor and others at work correlated higher with depression than didstressors such as role conflict, role ambiguity and lack of participation. In manystudies (Repetti, 1987; Reiche, 1982; Winnubst, Marcelissen and Kleber, 1982),the majority of correlations between social support and well-being measures aresignificant, and on the average around 0.30.

There are also a number of studies in which no or negligible correlations betweensupport and indicators of mental and physical health are found (.e.g Jayaratne andChess 1984). The lack of correlations is especially apparent when support, alongwith stressors, is entered into a regression analysis (e.g. Dignam, Barrera andWest, 1986; Constable and Russell, 1986). Furthermore, when differences arefound, support from one's superior generally seems to be more effective thansupport from one's colleagues in reducing stress reactions (see Constable andRussell, 1986; Repetti, 1987; Russell, Altmaier and van Velzen, 1987).

A noteworthy finding is that correlations are rarely found between support andobjectively assessable indicators of health, including blood pressure, cholesterollevel, heart rate, somatic diseases and absenteism, even when correlations withindexes of mental well-being are found in the same study. This lack of correlationshas been established in studies employing various occupational groups, includingnavy personnel, middle management in companies and nurses (Winnubst,Marcelissen and Kleber, 1982; Larocco and Jones, 1978; Reiche,1982; Kaufmanand Beer, 1986). In addition, negative relations between social support and well-being are frequently reported. In this study of legal professionals, Burke (1982)reported that 31,"/" of the correlates between social support and aspects ofoccupational stress were positive. Leiter and Meechan (1986) found indicationsthat a more cohesive and reciprocal social network was positively correlated toburn-out among the staff of a residential mental health rehabilitation centre.

Stress and affiliation reconsidered 159

To summarize the foregoing, it seems that, in general, social support correlatesmodestly with indicators of well-being and mental health, while correlations withmore objective health indicators are rarely found. Although these findings seem toindicate that social supportive relationships at work - especially those withsuperiors - are accompanied by better mental health, in some cases social supportbears no relationship, or even a negative one, to well-being. Before offeringexplanations for these findings, we will first describe the evidence for buffer effects.

BUFFER EFFECTS OF SUPPORT AT WORK

Despite the fact that buffer effects have been the main focus in research on socialsupport and occupational stress, the results on such effects are even moreinconclusive than those on direct effects. In most studies, several stressors andseveral stress reactions ('strains') have been examined. The number of possibleinteraction effects is obviously equal to the product of stressors and stress reactions.Despite the large number of interactions that are usually tested, a substantialnumber of studies reports no significant buffering effects, even with respect topsychological strains (see Sears, McGee, Serey and Graen, 1983; Leitner andMeechan, 1986; Larocco and Jones, 1978). In many other studies, the number ofsignificant effects hardly exceed what one would expect on the basis of chance.Thus, out of ten interactions that were examined, Constable and Russell (1986)found only one buffer effect. Similarly, in their study of burnout amongcorrectional officers, Dignam et al. (1986) reported just one effect out of 10 to besignificant. In his study of middle management, Reiche (1982), found only 6.8"/" ofthe interactions significant at the 5% level. Ganster, Fusilier and Mayes (1986)reported no more than three out of 24 interaction effects as significant, House(1985) 24 out of 140 effects, and Russell, Altmaier and Van Velzen (1987) three outof 30. Similar, or even more negative findings have been found in other studies (e.gBlau, 1981; Jajaratne and Chess, 1984).

Some studies have provided somewhat more support for the buffering hypothesis.In a study of Winnubst, Marcelissen and Kleber (1982), 11 of 60 buffer effects werefound, most with regard to feeling threatened and irritated. Health strains were lessoften counteracted by social support, however. LaRocco, House and French (1980)found more buffer effects than in most other studies. With regard to depression andanxiety, for instance, nearly half of the possible buffer effects were significant, andwith regard to somatic complaints a third of the effects were still significant.

In several studies, reversed buffering effects have been reported: in some casessocial support appears to increase the impact that stress has on well-being instead ofreducing the effect. For instance, a study by Kobasa and Puccetti (1983) involvingbusiness executives, showed that among high stress as well as low stress executives,the highest illness scores were found among those low in hardiness (a personalityvariable that includes a feeling of control, commitment and challenge), and high infamily support. As Kobasa and Puccetti noted, low hardiness executives may beencouraged by their families to allow themselves some self-pity and a preoccupa-tion with how they feel. Family support may take the executives low in hardinessaway from the work stresses and their resolution. In a study among nurses,Kaufman and Beehr (1986) found that for all of the sienificant interaction effects,

160 B.P. Buunk et al.

the relationship between sources of stress and stress reactions appeared to behigher among individuals who had access to strong social support systems thanamong individuals who lacked these systems.

Several explanations have been suggested regarding the frequently reportedfinding that social support is positively related to stress reactions. First, there is thepossibility of a spurious relationship, caused by a variable that correlates with socialsupport as well as with stressors. Certain jobs may act as a stressor andsimultaneously as a source of social support. Second, there is the possibil i ty ofsocial influence processes such as group-polarization (Meertens, 1980). Throughcontact with other people, the stressed person may become even more convincedthat things are wrong at his work, which may increase his or her feelings of distress.Third, stressed people often start looking for social support, which frequentlyresults in obtaining relatively more support. This does not necessarily result in adecrease of stress reactions, however. Fourth, other people may stimulateineffective coping behavior, such as the use of alcohol, or complaining andgrumbling, which may increase stress symptoms. Finally, social support may beperceived as undesirable. For instance, well-intended support can be interpreted asa threat to the self-esteem since it is perceived as implying inferiority in comparisonto the help-giver (Fisher, Nadler, and Whitcher-Alagna, 1983). Support may alsobe experienced as unpleasant since it induces feelings of indebtedness (Hatfield andSprecher, 1983).

ASSESSING STRESSORS INTERSUBJECTIVELY

In contrast to the pioneer study on occupational stress (Kahn et a\.,1964),intheoverwhelming majority of studies, the assessment of stressors is based solely uponthe perception of the repondent. At the same time it is supposed that such stressorsare the cause of strains. However, the opposite might also be true: the respondent'sperceptions of stressors may, in part, be colored by negative feelings that are theresult of, for instance, personal problems or a neurotic personality (Payne andJones, 1987). Furthermore, relationships between variables may be the result ofresponse-style or halo-effects. Respondents, for instance, may be prone to agree orto disagree with a statement, or perceive work related aspects either positively ornegatively due to a general positive or negative view of things.

In the present study, these problems were circumvented by intersubjectivelyassessing the extent to which a particular work setting is characterized by certainstressors. The study presents data from a research project on occupational stress,which included a sample of more than 2000 subjects from various organizations anddepartments. Because of the large number of subjects who participated in thisstudy, it was possible to determine the extent to which a certain work unit wascharacterized by the existence of particular stressors. While the stressors aredetermined intersubjectively, social support is conceived as a variable at theindividual level: in this study we are primarily interested in the subjectiveperception of the extent to which one's boss or one's colleagues are willing toprovide support. We assume that this perceived intention of support is primarilyrelated to the interpersonal relationships between the respondent and his or herboss or colleagues, and not to the climate in a department.

Stress and affiliation reconsidered 16l

The problem examined in the present study can be described as follows: is there astronger influence of subjectively experienced social support upon stress reactionsin those departments where certain stressors are clearly present, (according to theemployees working in that department), compared to departments in which(according to the general opinion of the employees) this specific stressor isperceived as less manifest. Only a few studies have followed an approach similar toours (c/. Repetti, 1987).

METHOD

SampleThe data for this study stems from a large scale longitudinal research project on

organizational stress. Subjects were more than 2000 employees of 21 Dutchcompanies from the Eastern part of the Netherlands. This study was a joint effortbetween the University of Nijmegen and the Industrial Health Service (BGD). Allsubjects completed a questionnaire on organizational stress. Additional resultsfrom this project have been reported elsewhere by Marcelissen, Winnubst, Buunkand de Wolff (1987), and by Winnubst, Buunk and Marcelissen (1988). For thepurpose of this article, 25 work units spread across 15 organizations were selected.The selection criterion was that every work unit had to consist of at least 25employees. In this manner, it was possible to have work units at our disposal thatwould be homogeneous enough, and at the same time large enough to makemeaningful comparisons possible. The selected sample consisted of 976 subjects.The majority were males (79.9%). The mean age was42 years. The educationalbackground of the subjects was quite diverse, and the selected organizations variedconsiderably. Included were industrial organizations, non-profit organizations, aswell as municipal organizations.

MeasuresThe questionnaire used was a modified version of the Organizational Stress

Questionnaire, (Van Dijkhuizen, 1984). The subjects also underwent a medicalexamination. The employee's systolic and diastolical blood pressure, (over)weightand cholesterol level were registered. However, not enough data concerningdiastolical blood pressure were available for this study.

The following scales measuring stressors in the work environment were used:

Role overload. A 9-item scale measuring the quantity of work, as well as thedegree to which the work exceeds one's capacities (alpha :0.77).

Role ambiguity. A scale containing four items, which referred to the extent towhich the role requirements were perceived as ambiguous, particularly the lack offeedback about one's skills and the lack of information about one's tasks (alpha :0.7r).

Role conflícl. A three items scale that mcasurcs the degree to which conflictingdemands exist in the work environment, particulariy contradictions in theperceived demands made by others (alpha : 0.70).

162 B.P. Buunk et al.

Inability to leave the work site. A two item index concerning the extent to whichone is l imited in leaving the work site (alpha : 0.53).

Lack of participation. A two item index dealing with one's lack of opportunitiesto participate in decisions regarding the way work needs to be conducted or theinabil ity to influence one's work pace (alpha : 0.66).

Meaninglessness of work. A three item scale that measures the extent to whichwork is negatively evaluated with regard to meaning, involvement and importance(alpha : 0 .68) .

Job future ambiguity. The amount of uncertainty a person has about thestability of the organization he is working for, measured by means of a four itemscale (alpha : 0.70).

Responsibility for others. A four item scale referring to the responsibility onehas for others (alpha : 0.70).

The following indexes were used to measure strains (stress reactions):

Job dissatisfaction. A three item index that measures the generally perceivedquality of the work (alpha : 0.62).

Spill over. Two questions concerning the difficulty one might have at home tomitigate work-related problems, and the frequency one talks about work at home.

Lack of hobbies. One item concerning the question of whether one lacks orengages in hobbies.

Cognitive anxiety. A scale that measures the extent to which one worries aboutthe future of the company, relationships with others, the fulfillment of conflictingorders, and the perceived abil ity to handle one's work (alpha : 0.58).

Health complaints. The number of complaints, such as dizzyness, rapidheartbeat, shortness of breath and lack of appetite.

Negative affect. A 13 item scale measuring the frequency of such negativeemotions as depression, anger and loneliness (alpha : 0.80).

Two measuresïor social support were included:

Supervisor support. A five item scale, referring to the possibility of communi-cation with the superior about work problems, the availability for supervisorsupport when things at work became difficult, the regard and esteem provided by

the supervisor, and the quality of the relationship with him (alpha : 0.83).

Coworker support. A similar scale, but referring to support by coworkers(a lpha : 0 .75) .

Stress and affiliation reconsidered 163

Data analysisTo test the tenability of the assumption that two work units differed with regard

to the extent to which certain stressors were manifest, we examined for everystressor by means of variance analysis (with work units as independent variable)whether a significant difference existed between the work units. This proved to betrue for every stressor. Next, for each stressor allthe work units were ranked, andthe four departments with the highest score and the four departments with thelowest score were selected. In this manner, we created (pertaining to everystressor) two groups of work units that could be called 'high stressor work units'and 'low stressor work units'. According to the buffering hypothesis of socialsupport, support from one's supervisor and one's colleagues should have a strongerstress reducing effect within the high stressor units in comparison to the low stressorunits. Stated otherwise: we should find some kind of interaction effect between astressor and social support, with stress reactions being the dependent variable.

Because it cannot be assumed that buffering effects upon the various stressreactions are independent from one another, a multivariate test was considerednecessary. We employed a test that is mn as a prelude to analyses of covariance,i.e. the test for the parallelism of regression l ines. It was tested whether theregression weights of the dependent variables (in this case: the stress reactions) onthe covariates (in this case: social support) were the same for every level of theindependent variable (here: high versus low stressor). Because a total of eightstressors and two variables of social support (supervisor and coworker support)were examined, the multivariate analyses were executed 16 times. When the testgenerates a multivariate significant effect, this means that social support in the highstressor group has a different effect on stress reactions than in the low stressorgroup. Next, it is possible to examine which stress reactions cause this difference atan univariate level. In order to interpret possible significant differences betweenthe regression lines, we have calculated (both with regard to the high and lowstressor organizations), the unstandardized regression weights (B) of the stressreactions on social support.

It is important to note that, in order to obtain a correct interpretation of theresults, social support was coded in such a way that a high score represents a lack of.social support. On the other hand, a high score on a strain indicated more stress.Thus, if the regression coefficient B is positive, this means that social support isrelated to a lack of strains. We assume the existence of a buffering effect when theregression coefficient B in the high stress departments is larger than in the lowstress departments, and also positive. If B is negative, this means that socialsupport increases the stress reaction, and more so to the extent that B is larger.

RESULTS

Table 1 presents the results of the multivariate and univariate tests, as described inthe previous sections. The interaction effects among the stress reactions and socialsupport are presented for each of the stressors. Where we found a significantunivariate interaction effect, the regression coefficients B are reported.

The results of the multivariate tests were significant for each of the stressors,indicating that, in general, the impact of social support is dependent upon the level

Table 1. Effect of social support upon strains within high and low stressor work units

À

fi

!

StÍ6sÍ Role overlod Role mbiguity Role @nllicl Inrbility to leve the w6k sire

multivuiatel) F(10,179F4.38r1 F(ro,224È2.69.' F(10,164F3.85r' F(1O177)=8.33'r

È

N(\

mivuiael)

F(df=1.1E8)

B mivuiatcl)!t6g

high F(F101) (d=1,233)

B mivuiaal)EEIM

bw high F(Fll6) (Fllg) (df=l,l73)

B uivuiacl)tw

BstÍss

Strrntb*

(Dd8)low

(F69)high F bw

(rlof) (dÊl,lE6) (íg)lhigh

(F96)

Supcrvlsr supportWdShlClplctmlSyrblicbl@dFswJob disrrirfrtionSpill ovqSpill ovc 2Lrkof lpHicCoFitivc dictyHolthmplainuNctÍivc rfrcct

.09

.37

.05ó.45rr

ló.09..13.71..l .E l

l6.tl...14

1.09

3.527.79..2.V32.6.o2

2.:U.9.10...78.51

1.19

2.r52.10r.08652.7.4orr.97

9.35rtIt.l6.r1.79t.2z

.54 .55

.21 .50

.05 .15

.2t .48

.52 .4t

.40 .44

-.M -.01-17 _40

2t35.. -8,83 ó.548.51.+ 1.61 -1.959.31.. 4.7E -.04

2t.7tt. .64 .49.73

5.52 . - .16 ,1411J0.. -.0t -.066.71 .25 .462.957.6'lt* -19 -45

nultivritel) F(10,179F4.7.. F(10,224ts3.06.. F(10,164F4.23.. F(r0,177Ft.96rr

Ccrorkcr supportWGiShtCholBErclSysolic blod p*ewcJob disrrisfrctbnSpill ovaSpill ovcr 2lek of hobbisCoFitive uictyHslth cmplaiilsNeguivc rfiet

.85 -.09

.35 .34

.28 .39

.18 .tl4

s.17. -2lJr____$.9:l9.91 -.r8 4.04l.6E2.93.04

4. t49.22.09. l I

2.63

4.55. -63.25 {9.022.36.94

10.ólxr .6ó -.034.82t .40 .00.34

1t-18*r .04 -.o214.46.* .21 .432.52.20

32.90*. -1E36 42.496.77. -.2/l -.t0

10.63r.u.61.. .30 .2t

.343.57

12.99.. -.05 .0Eó.48. .ló .354.24. .22 .366.51r .31 .44

.10

.05

.(B10.21+*13.67..13.48ri3.52

14.65r*

.00

Table 1 (continued)

Lrk of psticipcim Mcminglcms of wort Job fuur€ mtiguity Rcpouibility fc otlan

muldvrilal) F(l0,lt lFt.3l.. F(10.t42F2.95r. F(10,1ó5F5.3?ra F(10,19óF3.23..

uivuieel) B uivriarcl)stÍffi

F low high F(d=1,190) (ru2) G80) (ó=1.151)

B uivuiercl)llrs

hirh F(F54) (Gl,l?.f)

bw(íÉ99')

B uivriÍ.I) BsrÍ6s sE6s

low high F low hith(r44) 11*I\ (Clríb) (F99) (rl(f,)

Supcrvlaor suppoíWeitlÍCholaemlSysblicblmdFcswJob dircrtirfrtimSpill ovaSpill ovgr 2Lrkof bbtiaColnitivc mictyHalrhdrdrintsl{cartivcrfia

.11 .5t-,t'-----a

,tz---a

.47 -t.62

.53 .55

.02 .r3

10,79.. 13.(n 1.5ó.5t

1.47It.lsr. ,11 .37

.36

.6t7.3E.. -.0t_____.02t.a33tt.65r. .19 ,36

t.42l.tó.63

1430.. .4 .43l1.o4.r .39 .36332.38

5.03. .35 .1t.33

t3 t

l.7E1.874.Ezr

35.7E..t).69.r4.t1.t.715.0E..19.92.

t.29.53

-239.52

.774.22'.u

9.66r.z.r3

1027..t.57.32.15.77

mulrivritcl) F(10,r81F930r. F(10,142F332.. FO0,165F5.tt.. F(10,196F3.óói.

Co-rorkcr tlpportlvcitlttCblcrrrolSFblictilodpcsuc,obdisÍirfrtidSpill ovsSpill ova 2bkof ltobbyCognitivc uictyHalthcmÊriilrl{cgrrivc rffct

4.$ -8.75.53 .41.24 -'4)-.0?-_-04

.40

44.t3

.06 .50

--05__-ltz

.26 .lt

l . l2l.í)1.tl

13.ót.. .19 .4113.ó2.a J,Á-II5.32. .ll .05t . t3ó.94.. 24_________-n. t7

l .7 l

2.5r1.96.62'

34.4E..31.U..7.3E..2.69ó.3t..74.00

3.41.32

t.77.r2.9E

16.47...85.&.54.0?

l3.l3r..52

r.37B.I2t'

.0t

.xt17.u..1.523.15

10.52..

.42 -.t7

.53 -.16

q\(\q

\x

À

Q

Q

c;

(Jl

.35

Notc: Who the B's ac urdclirf,À a buffa cfret cxists.r p < . 0 5r r p < . 0 11) Multivqirte, rcspetivcly miyrirrr intastim cffat.

166 B.P. Buunk et al.

of the stressor. Univariate tests were executed next. Of these 160 tests, 66generated significant interaction effects. However, the expected buffering effectwas found in only about half of the cases (34 times). In these cases, social supporthad a stronger effect on stress reactions in th'e high stressor departments than in thelow stressor departments. As Table 1 makes clear, the impact of social supportdepends in part on the type of stressor. The stressor that most frequently showed abuffering effect was inabil ity to leave the work site. Similar effects were found withrespect to role overload, lack of participation and job future uncertainty.

The existence of buffering effects also appeared to depend upon the nature of thestrain. With regard to blood pressure, cholesterol level and psychosomaticcomplaints, buffering effects were only occasionally observed. Job dissatisfactionand negative affect were more frequently influenced by social support, but the mostconsistent buffering effects were found with respect to cognitive anxiety: the effectof supervisor and coworker support upon this variable was, in general, stronger inthe high stressor units than in the low stressor units. Similar, though somewhat lesspronounced, findings were established with regard to spil l-over of work problemsinto the situation at home, a variable that is, in fact, closely related to cognitiveanxiety.

Remarkably, with respect to half of the significant inieraction effects, nobuffering effect could be detected. Most of these cases even indicated a reversedbuffering effect: the effect of social support was stronger in the low stressor unitsthan in the high stressor units. This particularly applied to the stressorsresponsibility for others and lack of participation. Furthermore, it is noteworthythat the reversed buffering effects particularly appeared with respect to job

dissatisfaction. Eight cases showed a decrease in job dissatisfaction caused by socialsupport, especially in the low stressor units.

Even more in contrast to the buffering hypothesis was a phenomenon found in 14out of the 67 interaction effects, which could possibly best be defined as aboomerang effect. ln these cases, it appeared that social support aggravated thestress reactions or did not affect them at all in the high stressor units while, at thesame time, social support reduced the stress reactions in the low stressor units, oraggravated them to a lesser extent as was the case in the high stress departments.These effects were particularly observed with regard to role conflict.

To summarize, there were more buffering effects found in this study than in mostother studies. In general, the fact that so many of the interaction effects were

significant (although in many cases not in the expected direction) seems to underlinethe usefulness of assessing stressors intersubjectively. Buffer effects were particularlyobserved when the work was characterized by role overload, lack of possibilities ofleaving the work place, and job future ambiguity, and were seldom found withregard to the stressors role conflict and role ambiguity. In contrast, in many cases,such as with regard to the stressor role conflict, support seemed to aggravate stressto a considerable degree. As has been the case in other survey studies, objectiveindicators of health were hardly affected by support. On the other hand, cognitiveanxiety, and, to a lesser extent, negative affect appeared to be rather consistentlybuffered by social support.

DISCUSSION

The results of this study were generally consistent with previous findings in thisarea. In line with previous research, we also found buffering effects operating in a

168 B.P. Buunk et al.

and Wong, 1984; Shumaker and Brownell, 1984). It is striking that very l itt leresearch has been conducted that has focused on these kinds of social supportrelated processes. It seems both theoretically and practically worthwhile to gainmore insight into the conditions that cause positive and negative effects.

In addition to boomerang effects, reversed buffering effects were found: socialsupport reduced stress reactions, but this effect was stronger in the low stressorunits than in the high stressor units. This effect was found with respect to oneparticular stress reaction, namely job dissatisfaction. It is not clear how this effectcan be explained.

A final result that we would l ike to mention is that social support was hardlyrelated to physical health: physiological measures and health complaints were onlyoccasionally found to be affected by social support from one's colleagues and one'sboss. On the other hand, relatively consistent buffering effects were detected withregard to cognitive anxiety and spill-over. Worrying about the job, as well as beingunable to relax at home, were found to be reduced to a greater extent by socialsupport in the high stress units as compared to the low stress situation. It is obviousthat the finding that anxiety seems especially easily influenced by social support,can be interpreted within the framework of studies on fear and affiliation initiatedby Schachter (1959). Research in this area showed, for instance, that fear-provoking situations stimulate the seeking of the company of other individuals,especially when one is insecure about the nature of one's feelings. In addition, itappears that in such instances the company of others who are in a comparablesituation and show the same reactions is preferred (Wheeler 1974). Although socialcomparison of emotions was originally perceived as the most important process,Shaver and Klinnert (1982), in a crit ical discussion of Schachter's work, suggestedthat the motivation behind affi l iation with respect to fear-provoking situations istwofold: on the one hand, people show a need for more insight into the nature ofthe threatening situation (cognitive motive), while on the other hand they attemptto reduce feelings of fear (affective motive). Both needs, contrary to Schachter'sassumption, cannot only be fulf i l led by the company of fellow sufferers, but also bythe interaction with emotionally significant persons.

The extent to which such processes actually occur with respect to the socialsupport related reduction of occupational stress is not totally clear. Most of theresearch concerning organizational stress is too general to discover the cognitiveand affective interpersonal processes that could be of importance in this context(Buunk, 1988). In addition, l i tt le is known about the conditions, events andemotions in organizations that provoke the need for social support - to whom doesone turn to gain support, what occurs in such interactions, what are theconsequences of these interactions, what keeps people from seeking support, andto what extent can mere companionship have a stress-reducing effect. In futureresearch, these issues deserve more attention to obtain more insight in the way inwhich interpersonal relationships can help in alleviating the potential negativeconsequences of work stress, a form of stress that constitutes a serious threat towell-being and health in contemporary society.

Stress and affiliation reconsidered 167

minority of various stressful situations within organizations, although it seemedthat the manifestation of those effects was less random than the l iterature suggests.

First, buffering effects were particularly registered with respect to certain typesof stressors. More specifically, social support seemed to have relatively strongstress reducing effects in departments that were characterized by role overload,lack of possibil i t ies of leaving the work site, lack of participation, and job futureambiguity. This in contrast to departments which barely manifested these stressors,or did not manifest them at all. The difference between high stress and low stressunits was less manifest or non-existent when these units differed in the degree ofrole conflict and role ambiguity. At present, it is sti l l unclear what differentiatesthese stressors. It is possible that those stressors upon which social support has thestrongest impact are especially relevant among blue collar workers. It is alsopossible that these stressors induce different emotions than stressors where socialsupport does not have positive effects. For instance, experimental research suggeststhat if stress is caused by failure or pain, the presence of others may have a stressreducing effect, whil : stress which is caused by embarrassing situations isaggravated by the presence of others (Cottrell and Apply, 1977). Thus, a crucialintervening variable that determines the effect of social support may be the degreeto which stress induces feelings that may evoke social disapproval.

The results referred to as boomerang effects of social support represent anillustration of the negative effects of support. In a substantial number of cases socialsupport seemed to aggravate stress reactions in the high stressor units, while itsimultaneously reduced stress reactions in the low stressor units. Stated differently,when stressful circumstances arise (and individuals may have a need for support),social support seems to have negative effects in some cases. In the foregoing, wealready presented several explanations for the general negative effects of socialsupport, some of which seem applicable in this context. For instance, grouppolarization (Meertens, 1980) may cause an exacerbation of the seriousness of thesituation and a stimulation of ineffective coping behaviors. Thus, through a processof information exchange and social comparison, persons may become moreconvinced of the fact that their situation is bad, and develop more negative feelingssuch as anger or fear. Furthermore, others may not only influence the perception ofstressors, and thereby the stress reactions, but may also have a direct influenceupon these reactions by acting as models showing certain symptoms that areimitated. The classical experiments of Schachter and Singer (1962) indicate, in fact,that artificially induced arousal makes it more likely that certain reactions areimitated from another person in the same situation. In a similar vein, Pennebakerand his colleagues (Skelton and Pennebaker, 1982; Pennebaker and Britt ingham,1982) have, in a series of experiments, shown that various physical symptoms,including scratching, coughing, and flu symptoms are easily adopted from others.Skelton and Pennebaker (1982) have suggested that persons under stress maydevelop hypotheses on having a certain disease and begin to look inside themselvesfor symptoms, thus, in a process of self-fulfilling hypotheses, they develop a certainpsychosomatic disease.

As suggested earlier, it is also possible that social support itself can be perceivedin a negative manner. This might, for instance, occur when receiving support isexperienced as a threat to self esteem, implies an inferior position, or conflicts withvalues with regard to autonomy and self control (Antonucci, 1985; Shin, Lehman

Stress and affiliation reconsidered 169

REFERENCES

Antonucci, T.C. (1985). 'Social support: theoretical advances, recent f indings and pressingissues'. In: Sarason, I.G. and Sarason, B.R. (Eds) Social Support: Theory, Research andApplications, Martinus Nijhoff, Boston.

Barrera, M. (1986). 'Dist inct ions between social support concepts, measures and models' ,American Journal of Community Psychology,l4(4)t 413445.

Beehr, T.A. (1976). 'Perceived situational moderators of the relat ionship betweensubjective role ambiguity and role strain', Journal of Applied Psychology,6l:35-40.

Bi l l ings, A.G. and Moos, R.H. (1982). 'Work stress and the stress-buffering roles of workand family resources', Journal of Occupational Behaviour,3z 215-232.

Blau, G. (1981). 'An empir ical investigation of job stress, social support, service length, andjob strain', Organizational Behavior and Human Performance,2Tz 279-302.

Brenner, S., Sorbom, D. and Wall ius, E. (1985). 'The stress chain: a longitudinalconfirmatory study of teacher stress, coping and social support', Journal of OccupationalPsychology,58: 1-13.

Burke, R.J. (1982). ' Impact of occupational demands on nonwork experiences', Journal ofPsychology, ll2 (2): 195-211.

Buunk, B. (1988). Companionship, intimacy and support in organizations: a microanalysis ofthe stress-reducing functions of social interactions. Paper presented at the 4th InternationalConference on Personal Relationships. Vancouver, Canada, July.

Buunk, B. (1989). 'Affiliation and helping within organizations: a critical analysis of the roleof social support with regard to occupational stress'. In: Hewstone, M. and Stroebe W.(Eds) European Review of Social Psychology, vol. 1, John Wiley, Chicester.

Buunk, B.P. and Janssen, P.P.M. (1987). 'Sociale ondersteuning in de thuissituatie enpsychosociale stress op het werk' (Social support at home and psychosocial stress at work),Gedrag en Gezondheid,l5z 147-154.

Caplan, R.D., Cobb, S., French, J.R.P., van Harrison, R. and Pinneau, S.R. (1975). "ÍobDemands and Worker Health, Main Effects and Occupational Differences. U.S.Department of Health, Education and Welfare. National Institute for Occupational Safetyand Health, Washington, D.C.

Cohen, S. and Wil ls, T.A. (1985). 'Stress, social support and the buffering hypothesis' ,Psychological Bulletin, 98 (2): 310-357.

Constable, J.F. and Russell , D.W. (1986). 'The effect of social support and the workenvironment upon burnout among nurses', Journal of Human Stress, llt 2W27.

Cottrel l , N.B. and Epley, S.W. (1977). 'Aff i l iat ion, social comparison, and social lymediated stress reduction'. In: Suls, J.M. and Miller R.L. (Eds) Social ComparisonProcesses. Wiley, New York.

Dignam, J.T., Barrera, M. and West, S.G. (1986). 'Occupational stress, social support andburnout among correctional officers', American Journal of Community Psychology,14:177-193.

Fisher, J.D., Nadler, A. and Witcher-Alagna, S. (1982). 'Recipient reactions to aid' ,Psychological Bulletin, 9lz 27-54.

French, J.R.P. and Kahn, R.L. (1962). 'A programmatic approach to studying the industr ialenvironment and mental health', Journal of Social Issaes, l8 (3)z 1.4'1 .

Ganster, D.C., Fusi l ier, M.R. and Mayes, B.T. (1986). 'Role of social support in theexperience of stress at work', Journal of Applied Psychology,Tlz 102-110.

Hatfield, E. and Sprecher, S. (1983). 'Equity theory and recipients reactions to help'. In:Fisher, J.D., Nadler, A. and dePaulo, B.M. (Eds) New Directions in Helping Behavior. LlI3-141, Academic Press, New York.

House, J.S. (1981). Work Stess and Social Support. Addison Wesley, Reading, MA.Jayaratne, S. and Chess, W.A. (1984). 'The effect of emotional support on perceived job

stress and strain', Journal of Applied Behavioral Science,2lz l4l-153.Kahn, R.L. (1981). Work and Health, Wiley, New York.Kahn, R.L . , Wol fe , D .M. , Qu in , R.P. , Snoek, J .D. and Rosentha l , R .A. (1964) .

Organizational Stress.' Studies in Role Conflict and Ambiguiry, Wiley, New York.

I70 B.P. Buunk et al.

Kaufman, G.M. and Beehr , T .A . (1986) . ' In te rac t ions be tween job s t ressors and soc ia lsupport: some counterintuitive results', Journal of Applied Psychology, Tl (3): 522-526.

Kessler, R.C., Price, R.H. and Workman, C.B. (1985). 'social factors in psychopathology:stress, social support and coping processes', Annual Review of Psychology,36: 531-572.

Kobasa, S.C. and Puccett i , M.C. (1983). 'Personali ty and social resources in stressresistance', Journal of Personaliry and Social Psychology,45: 839-850.

LaRocco, J.M. and Jones, A.P. (1978). 'Co-worker and leader support as moderators ofstress-strain relationships in work situations', Journal of Applied Psychology , 63 (5): 629-634.

LaRocco, J.M., House, J.S. and French, J.P.R. (1980). 'Social support, occuparional stress,and health', Journal of Health and Social Behavior,2l:.202-218.

Leiter, M.P. and Meechan, K.A. (1986). 'Role structure and burnout in the f ield of humanservices', Journal of Applied Behavioral Science,22 (l): 47-52.

Marce l i ssen, F .G.H. , Winnubst , J .A .M. , Buunk, B . and de Wol f f , Ch.J . (1988) . 'Soc ia lsupport and occupational stress: a causal analysis', Social Science and Medicine,26: 365-J I J .

Meertens, R.W. (1980). Groepspolarisatie, Yan Loghum Slaterus, Deventer.Payne, R. and Fletcher, B. (1983). 'Job demands, supports, and constraints as predictors of

psychological strain among school teachers' , Journal of Vocational Behavior, 222 136-147 .Payne, R. and Jones, J.G. (1987). 'Measurement and methodological issues in social

support'. In: Kasl S.V. and Cooper C.L. (Eds) Stress and Health: issues in ResearchMethodology. Wiley, Chicester.

Pennebaker, J.W. and Brit t ingham, G.L. (1982). 'Environmental and sensory cues affect ingthe perception of physical symptoms'. In: Baum A. and Singer J.E. (Eds) Advances inEnvironmental Psychology. Erlbaum, Hillsdale, NJ.

Reiche, H.J.M.K.I. (1982). Stress aan het Werk. Swets and Zeit l inger, Lisse.Repett i , R.L. (1987). ' Individual and common components of the social environment at

work and psychological well-being', Journal of Personality and Social Psychology, 52t710-721.

Rook, K. (1984). 'Research on social support, lonel iness and social isolat ion', Review ofPerso nal ity and Social Psycholo gy, 5z 234-264.

Russell , D.W., Altmaier, E. and Van Velzen, D. (1987). 'Job related stress, social support,and burnout among classroom teachers', Journal of Applied Psychology,T3 (2):269-274.

Sarason, LG. and Sarason, B.R. (1985). 'Social support - insights from assessment andexperimentation'. In: Sarason, I.G. and Sarson B.R. (Eds) Social Support: Theory,Research and Applications, Martinus Nijhoff, Boston.

Sarason, I .G. and Sarason, B.R. (1986). 'Experimental ly provided social support ' , Journalof Personality and Social Psychology,50z 1222-1225.

Schachter, S. (1959). The Psychology of Affiliation. Stanford University Press, Palo Alto,CA.

Schachter, S. and Singer, J.E. (1962). 'Cognit ive, social and physiological determinants ofemotional sïate', Psychological Review, 692 379-399.

Seers, A., McGee, G.W., Serey, T.T., and Graen, G.B. (1983). 'The interaction of jobstress and social support: a strong inference investigation', Academy of ManagementJournal, 26 (2)z 273284.

Sell , M. van, Brief, A.P. and Schuler, R.J. (198i). 'Role confl ict and ambiguity: integrationof the literature and directions for future research'. Human Relations.34: 43-72.

Shaver, P. and Klinnert, M. (1982). 'Schachter 's theories of aff i l iat ion and emotion:implications of developmental research', Review of Personality and Social Psychology,3za t - I L -

Shinn, M., Lehmann, S. and Wong, N.W. (1984).'Social interaction and social support',Journal of Social Issues, 40:. 55-76.

Shumnaker, S.A. and Brownell, A. (1984) 'Toward a theory of social support: closingconceptual gaps'. Journal of Social lssues, 40: 11-36.

van Dijkhuizen, N. (1980). From Stressors to Strains. Swets and Zeitlinger, Lisse.van Dijkhuizen, N. (1984). 'Questionnaire for organizational stress'. In: Cullen J. and

Siegrist J. (Eds) Breakdown in Human Adaptation to Stress, vol. l, Psychological and

Stress and ffiliation reconsidered I7l

Sociological Parameters for Studies of Breakdown in Human Adaptation. MartinusNijhoff, Boston, MA.

Wheeler, L. (1974). 'Social comparison and selective affiliation'. In: Huston T. (Ed.)Foundations of Interpersonal Attraction. Academic'Press, New York.

Winnubst, J.A.M., Buunk, B. and Marcelissen, F.G.H. (1988). 'social support and stress:p_erspectives and processes'. In: Fisher, S. and Reason J. (Eds) Handbook of Life stress,Cognition and Health, Wiley, New York.

winnubst, J.A.M., Marcelissen, F.G.H. and Kleber, R.J. (1982). 'Effects of social supportin the stressor-strain relationship: a Dutch sample', Social Science and Medicine,16;475-482.

Zander, A. and Quinn, R. (1962). 'The social eníironment and mental health: a review ofpast research at the Institute for Social Research', Journal of Social lssaes, lE (3): 48-67.