Sei sulla pagina 1di 12

Journal of Occupational Health Psychology Copyright 2003 by the Educational Publishing Foundation

2003, Vol. 8, No. 3, 220 231 1076-8998/03/$12.00 DOI: 10.1037/1076-8998.8.3.220

The Enigma of Social Support and Occupational Stress:


Source Congruence and Gender Role Effects
Terry A. Beehr and Suzanne J. Farmer Sharon Glazer
Central Michigan University San Jose State University

David M. Gudanowski and Vandana Nadig Nair


Central Michigan University

Research on the potential ameliorating effects of social support on occupational stress produces
weak, inconsistent, and even contradictory results. This study of 117 employees, mostly from a
southern U.S. hospital supply company, examined potential moderators that were theorized might
reduce the confusion: source congruence (congruence between sources of the stressor and of
social support) and gender role. Congruence between the sources of stressors and of social support
appeared to make little difference in determining the moderating or buffering effect of social
support on the relationship between stressors and strain. Gender role, however, may moderate the
relationship between social support and individual strains such that more feminine people react
more strongly and positively to social support than more masculine people do.

Occupational stress occurs when characteristics of main effect of social support on strains, and it exam-
the work environment (stressors) affect an employ- ines the concept of source congruence as a possible
ees health and welfare adversely (strain; e.g., Kahn explanation of the moderating effects of social sup-
& Byosiere, 1992), but interpersonal contact with port on the relationship between occupational stres-
supportive others has long been recommended as a sors and individual strains.
means of alleviating some of this strain. There are Social support sometimes has been found to act as
two primary ways that social support is thought to a moderator of the stressorstrain relationship (Kahn
affect strain. It can have a direct effect in which & Byosiere 1992; Viswesvaran, Sanchez, & Fisher,
support directly reduces strain, perhaps by calming 1999). This was once the dominant hypothesis re-
the person, and it can have a buffering (type of garding social support and occupational stress (Gan-
moderating) effect in which it weakens the relation- ster, Fusillier, & Mayes, 1986), but enthusiasm for it
ship between stressors and strains. Research results has waned in recent years because of inconsistent
for both of these effects have been weak and incon- results. The moderator or buffering effect is repre-
sistent, however. The present study looks at two sented by a statistical interaction in which those who
separate models, one investigating main effects and receive more social support have a weaker positive
one investigating moderating effects of social sup- stressorstrain relationship than those with less social
port. It looks at gender as a possible moderator of the support. Overall, the research findings of buffering or
moderating effects have been rather tenuous, incon-
sistent, and unclear (Beehr, 1995; ODriscoll &
Terry A. Beehr, Suzanne J. Farmer, David M. Gu- Dewe, 2001). Beehr and Glazer (2001) conjectured
danowski, and Vandana Nadig Nair, Department of Psy- that these inconsistencies may be a result of the way
chology, Central Michigan University; Sharon Glazer, De- social support is studied and the context in which it is
partment of Psychology, San Jose State University.
Suzanne J. Farmer is now at Celera Genomics, Rockville, studied; their article primarily addresses cultural con-
Maryland; David M. Gudanowski is now at Personnel De- texts. Thus prior research and discussions suggest
cisions International in Brussels, Belgium; and Vandana further examination of interaction effects, but with a
Nadig Nair is now in Bangalore, India. need for research to be guided by reasoned theory
We thank Steve Jex, Adam Johnson, and Thomas Taber
for their helpful feedback on earlier versions of the article. that might predict conditions under which buffering
Some of these data were previously reported at a meeting of occurs or does not occur.
the Society for Industrial and Organizational Psychology in Overall, in spite of the common belief, intuitive
St. Louis, Missouri, in April 1997. appeal, and promise of social support as a treatment
Correspondence concerning this article should be ad-
dressed to Terry A. Beehr, Department of Psychology,
for occupational stress, its success and the specific
Central Michigan University, Mt. Pleasant, Michigan ways in which it might work have remained a mys-
48859-0001. E-mail: Beehr1TA@cmich.edu tery. In the present study, we examined both situa-
220
SOCIAL SUPPORT 221

tional (congruence) and personal (gender) factors. might be responsible for additional strain rather than
We examined (a) the relation between social support reduced strain, causing reverse buffering.
from multiple sources (i.e., supervisor, coworker) Regarding the source of stressors and supports, the
and psychological strain, (b) the relation between first hypothesis was simple: As with any stressor, the
stressors and strains when the source of the stressor is stressors emanating from the supervisor must be re-
congruent versus incongruent with the source of the lated to strains. Research has not previously paid
support, and (c) gender roles as a potential moderator much attention to sources of stressors, but one stres-
of the relationship between support and strains. sor scale that for some of the items referenced the
supervisor as the source of the stressor was a role
ambiguity scale (Beehr, Walsh, & Taber, 1976). Role
Source Congruence and the Moderating Effect ambiguity is a stressor for which it is especially likely
of Social Support that the supervisor could be the source. Part of a
supervisors job is to give directions to subordinates,
Failure to find consistent buffering or moderating
and such directions can have varying degrees of
effects for social support has led to conjecture about
clarity. A second commonly studied stressor for
the nature or very existence of social support buffer-
which subordinates often can attribute causation to
ing effects. Whereas some research has found the
their supervisors is workload, and therefore we also
buffering effect (e.g., Beehr, King, & King, 1990;
examined this stressor in the study. By definition,
Lim, 1997), other research has failed to find it (e.g.,
stressors should be related to strains, but because we
Ganster et al., 1986; King, King, Fairbank, Keane, &
Adams, 1998; La Rocco & Jones, 1978; Wade & had to rewrite the stressor items to make them refer-
Kendler, 2000), and still other studies have found a ence the supervisor as the source of the stressor, we
reverse buffering effect (e.g.., Glaser, Tatum, needed to confirm that these revised stressor mea-
Nebeker, Sorenson, & Aiello, 1999; Kaufmann & sures would be related to strains.
Beehr, 1986). A reverse buffering effect occurs when
Hypothesis 1: Role ambiguity and workload
social support strengthens rather than weakens the
originating from the supervisor will be posi-
positive relation between stressors and strains.
Viswesvaran et al.s (1999) meta-analysis argued tively related to strains.
that theoretical attention needs to be given to the
In the present study, source congruence exists
types of social support expected to moderate effects
when the person or source of an employees social
of specific types of stressors. As a result of the
support is the same as the person or source of the
perplexing findings, Blau (1981) had earlier sug-
stressor. Social supportstressor congruence hypoth-
gested that sources of support may need to be inde-
pendent of sources of stressors for support to be eses have been proposed (e.g., Blau, 1981; Kaufmann
effective. For example, how effective would the sup- & Beehr, 1986) but not tested in previous research.
port from a supervisor be, if that particular supervisor As suggested by Blau (1981), it may be better for
was also the source of the stressor? Kaufmann and sources of support to be different from the sources of
Beehr (1986) also suggested that future research the stressor. Otherwise, reverse buffering might oc-
should examine this congruence between source of cur. If stressors are due to the supervisor, having him
the stressor and source of support as an explanation or her come near for any reason (e.g., even to offer
for reverse buffering effects. Although none of these support) may be aversive, because the supervisor
researchers suggested a theory to explain such a becomes a stimulus reminding the subordinate of the
phenomenon, one explanation comes from cognitive stress. Regarding source congruence, therefore, pri-
dissonance theory (Festinger, 1957; Harmon-Jones, marily following the reasoning of Blau (1981) and of
1999). When the sources of the stressor and of the Kaufmann and Beehr (1986):
support are both the supervisor (as in the present
study), the employee may be left with two conflicting Hypothesis 2a: When the source support is dif-
cognitions about the supervisor (e.g., my supervisor ferent from the source of the stressor (source
imposes stressors on me and my supervisor offers incongruence), the relationship between stres-
support to me). According to dissonance theory, sors and strains will be moderated by social
such dissonant cognitions would be a source of dis- support such that the relation between stressors
comfort or tension, adding to the persons strain. and strains will be stronger for low social sup-
Thus congruence between source of stressor and port (i.e., a buffering effect) than high social
source of support (i.e., the source is the same person) support.
222 BEEHR, FARMER, GLAZER, GUDANOWSKI, AND NAIR

Hypothesis 2b: When the sources of the stressor nine gender role displays characteristics that empha-
and support are the same (source congruence), size warmth, nurturance, and compassion (Bem,
the relationship between stressors and strains 1974). Although it is generally believed that women
will be moderated by social support such that tend to assume traditional feminine characteristics
the relation between stressors and strains will be and men assume more masculine characteristics, it is
stronger for high social support (i.e., a reverse more likely that these feminine and masculine role
buffering effect) than low social support. characteristics vary across individuals within each
gender. The gender roles can then account for the
variation in the use of social support and its relation
Gender Role and the Main Effect of Social with strains, but previous research has not addressed
Support on Strain this possibility. On the basis of this rationale, the
following hypothesis was proposed:
Literature reviews (e.g., Beehr, 1995; Cohen &
Wills, 1985; Kahn & Byosiere, 1992; Viswesvaran et Hypothesis 3: The relationship between social
al., 1999) and large-scale empirical studies (e.g., support and strains will be moderated by gender
Caplan, Cobb, French, Harrison, & Pinneau, 1975) role such that the negative relation between sup-
have examined direct effects of support on strains. port and strains will be stronger for people as-
There is generally a negative relation between social suming feminine roles (high femininity) than for
support and measures of psychological strain, but the those assuming masculine roles (low femininity).
correlations are often only in the teens or near .20
(e.g., meta-analysis by Viswesvaran et al., 1999). It The present study, therefore, advanced our knowl-
seems quite possible that the relationship might be edge about social support and occupational stress by
stronger for some people than others, and the present testing proposals about source congruence and buff-
study examined gender role as a possible explanation. ering effects of social support, as well as about gen-
Gender has been found sometimes to influence the der role and main effects of social support.
relation between social support and strains (Billings
& Moos, 1982; Greenglass & Burke, 1988; Perrewe Method
& Carlson, 2002), although this has not been a wide-
spread finding (e.g., Vermeulen & Mustard, 2000). Sample
For example, Greenglass and Burke found that social
support was negatively related to burnout in women Questionnaires were distributed to 270 employees of ad-
but not men. They suggested that women may use ministrative, customer service, distribution, and manage-
ment departments in a southern hospital supply company.
social support more effectively than men Because Of those, 102 returned completed questionnaires. Because
asking for help is congruent with traditional feminine this number was smaller than expected and desired, through
gender role prescriptions and noncongruent with personal contacts we were able to distribute 15 more ques-
those prescribed by the masculine gender role tionnaires each to both an aircraft manufacturer and a phar-
maceutical company in the same city, with 9 and 6 re-
(Greenglass & Burke, 1988, p. 226). sponses, respectively. The total response rate was 39%.
Based primarily on Greenglass and Burkes (1988) Among the total 117 participants, 57% were female; 85%
reasoning, a logical alternative explanation to ones were White, 9% Black, 3% American Indian, 1% Hispanic,
gender as a characteristic influencing the effects of and 1% Asian American; the average age was 38.8 years.
social support is ones gender role. Prior studies The average number of hours worked per week was 45.6 hr.
Of the sample, 58% were married, and 46% of respondents
regarding gender and social support may have been had no dependent children at home. Finally, 71% had high
based on stereotypical viewpoints regarding tradi- school diplomas, 21% had bachelors degrees, and 3% had
tional behaviors assumed to be commonly exhibited graduate degrees.
by men and women, but gender roles rather than
biological sex may be responsible for differences
Measures
found. In fact, reasoning usually given for a possible
effect of gender is based directly on characteristics of Social support from supervisors and from coworkers.
the gender role rather than on biological sex (e.g., Two social support measures were used: social support from
Greenglass & Burke, 1988). These gender roles are the supervisor and social support from the coworkers. A
not, of course, perfectly matched to actual gender. modified version of the social support scale developed by
Beehr et al. (1990) and more recently used in Chen, Popo-
The masculine gender role has generally been de- vich, and Kogan (1999) and in Stephens and Long (2000)
scribed as focusing on instrumentality, competence, measured social support that specifically assessed the con-
assertiveness, and independence, whereas the femi- tents of communication between the respondent and other
SOCIAL SUPPORT 223

people at work. The original scale was constructed specif- person to do? and How often does your supervisor give
ically for a sample of nurses, and therefore, the content of you enough time to do what is expected of you on your
five items needed to be modified to make the items appli- job? Response options for all five items ranged from 1
cable to the different job types in the present sample. For (never) to 5 (extremely often).
example, the item We talk about problems in working with Strains. Three potential stress outcomes or strains were
doctors was changed to We talk about problems in work- measured: job dissatisfaction, anxiety, and psychological
ing with bosses (contact people in the company informed strain. Job dissatisfaction (M 2.93, SD 1.67, .90)
us this would be the appropriate terminology). There were was measured by reverse scoring the three-item job satis-
two social support scales, one asking about communications faction measure of the Michigan Organizational Assessment
with supervisors and one asking about communications with Scale (Cammann, Fichman, Jenkins, & Klesh, 1979). An
coworkers, used to measure the three contents of commu- overall job satisfaction index was created by averaging the
nication (i.e., positive, negative, and non-job-related com- responses to the three items (two items were reverse
munications). The original scale only measured social sup- scored). Response options ranged from 1 (strongly dis-
port from supervisors, but subsequent studies (Beehr, Jex, agree) to 7 (strongly agree). High scores indicated
Stacy, & Murray, 2000; Fenlason & Beehr, 1994) had more dissatisfaction.
adapted it for coworkers simply by changing the referent Anxiety (M 2.25, SD 0.73, .89) was assessed
person, and that was also done in the present study. A main using a modified version of the state scale of Spielbergers
focus of the present study was the comparison of results for (1979) StateTrait Personality Inventory that contained 10
supervisor support versus coworker support, and therefore items asking the respondents to describe how they felt at
the three supervisor support subscales and the three co- work during the past 30 days.
worker subscales were collapsed into two scales by averag- Psychological strain (M 1.95, SD 0.59, .90) was
ing the mean scores of the appropriate subscales. Reliabili- measured by the 12-item version of the General Health
ties for these scales were .92 for the supervisor support scale Questionnaire (GHQ; Banks et al., 1980; Goldberg, 1972).
(M 2.36, SD 0.78) and .85 for the coworker support Respondents were asked to respond to symptoms that oc-
scale (M 3.16, SD 0.60). The response scale ranged curred over the last 4 6 weeks, and all items were also
from 1 (strongly disagree) to 5 (strongly agree). After preceded by the phrase Have you recently . . . Some
reverse scoring the items that are worded negatively, higher items included lost much sleep over worry? been able to
scores were indicative of more support. The mean for the face up to your problems? and been feeling unhappy or
supervisor support scale was a little below the scales mid- depressed? A composite index was formed by calculating
point, which may indicate a sample with somewhat low
the mean of the items, which were answered on a 4-point
communications support from the supervisor.
Likert-type scale.
Stressors from supervisors. Two stressors were mea-
Gender role characteristics. Gender role characteristics
sured in the study: workload due to the supervisor and role
were measured with a modified version of the Bem Sex-
ambiguity emanating from the supervisor. Because stressor
Role Inventory (BSRI; Bem, 1974). Only the Masculinity
measures do not usually focus on the source of the stressor,
and Femininity scales, each containing 20 personality char-
the present study needed to adapt items from previ-
acteristics, were used in the present study. Using a 7-point
ous scales.
Role ambiguity emanating from the supervisor (M Likert scale anchored at 1 (never or almost never true) and
2.92, SD 1.22, .76) was measured by a modified 7 (always or almost always true), respondents were asked to
four-item Likert-type scale reported by Beehr et al. (1976). indicate how well each of the characteristics describes him-
Two items, which measured whether supervisors made sure self or herself. Masculinity and Femininity subscale scores
people had clear goals to achieve and how people should do were computed as the mean rating of all masculine and
their work, were reversed scored in the present study and feminine items, respectively.
were identical to those used in the original scale. The other Six graduate students in industrial and organizational
two items needed to be modified to indicate the supervisor psychology, acting as subject matter experts, were provided
as the source of the ambiguity. These two items asked if with a definition of social support and were asked to rate the
people knew what the supervisors performance standards degree to which each of the 40 personality characteristics
were for them and whether it was clear what the supervisor from the BSRI were descriptive of people who would like to
expected of them (reverse scored). The response options receive some type of social support. For example, some of
ranged from 1 (strongly disagree) to 7 (strongly agree). the items these graduate students rated included Affection-
Workload due to the supervisor (M 3.03, SD 0.85, ate people like receiving social support and Independent
.85) was measured by a five-item Likert-type scale that people like receiving social support (affectionate and in-
consisted of three items modified from a workload scale by dependent were potentially self-descriptive items on the
Spector, Dwyer, and Jex (1988) and two items modified gender role inventory). Response options ranged from 1
from a role overload scale by Beehr et al. (1976). As with (strongly disagree) to 5 (strongly agree). The characteristics
the previous scale, the goal was to assess the extent to which associated with the words in the feminine role were judged
the source of the stressor was the supervisor. The items in as more likely to be related to liking to receive social
the workload scale by Spector et al. (1988) were modified as support than the masculine characteristics were (M 4.18
follows: How often does your supervisor require you to vs. M 3.05), t(5) 5.25, p .01. This preliminary
work very hard? How often does your supervisor require analysis was a check that was consistent with our assump-
you to work very fast? and How often does your super- tion that people with more feminine and less masculine
visor give you more work than you can do well? Likewise, characteristics were more likely to want or desire social
the two items from the scale by Beehr et al. were modified support. The Masculine subscale was reverse scored, and
to reference the supervisor. The two items were How often then the mean of the Feminine and Masculine subscales was
does your supervisor give you too much work for one calculated as the gender role score (M 4.16, SD 0.51).
224 BEEHR, FARMER, GLAZER, GUDANOWSKI, AND NAIR

A higher score, therefore, indicated more femininity and the support both originating from the same source,
less masculinity. The reliability was .81. the supervisor). The hypotheses predicted that the
two interactions in the incongruent situations would
Results be instances of buffering effects, whereas the two
congruent interactions would be reverse buffering.
Hypothesis 1 was strongly supported. All six cor-
To examine the nature of each interaction, we divided
relations between the two stressors and the three
the sample at the median on the appropriate social
strains were positive, significant (p .01), and strong
support variable and calculated bivariate regressions
or at least moderately strong (see Table 1). The
of the strain on the stressor within each subsample.
strongest correlation was between workload and psy-
Figures 1 and 2 show that in each pair of regressions,
chological strain (r .50), and the weakest was
the beta coefficients for the low social support con-
between ambiguity and anxiety (r .25). Two out-
ditions were more positive than the betas for the high
comes, anxiety and psychological strain, were
social support conditions. Thus all four interactions
strongly related to each other (r .76), but their
relationships with the stressors were not identical. In were buffering effects. This was true regardless of the
fact, their correlations with ambiguity were signifi- stressor (workload vs. ambiguity) and regardless of
cantly different from each other, t(114) 3.70, p the strain (anxiety vs. psychological strain). Because
.01. Therefore they give somewhat different informa- the nature of the interaction effects was the same (all
tion about their relationships with stressors. buffering, and no reverse buffering), the source con-
Regarding Hypotheses 2a and 2b, three of the six gruence hypothesis was not supported.
interactions between workload and support predict- Regarding Hypothesis 3, one of six interactions
ing strains were significant (see Table 2), and one of regarding gender role was significant at p .05 (see
the six interactions between ambiguity and support Table 4). Supervisor support and gender role inter-
was significant (see Table 3), controlling for age and acted to predict job dissatisfaction (R2 change .03).
gender. Workload interacted with both supervisor In addition, the interaction between gender role and
support (R2 change .03) and coworker support (R2 supervisor support predicting psychological strain
change .06) to predict anxiety, and workload in- was marginally significant (R2 change .03, p
teracted with coworker support to predict psycholog- .10). The nature of these two interactions was exam-
ical strain (R2 change .03). Ambiguity interacted ined by dividing the sample at the median on the
with supervisor support to predict anxiety (R2 moderator variable (gender role) and calculating bi-
change .01). These four significant interactions variate regressions of the strain on social support for
were examined further. Two of them were incongru- each subsample (see Figure 3). The beta coefficients
ent situations (the level of stressor originating from for the high femininity subsample were more nega-
the supervisor, combined with the level of support tive than for the low femininity subsample in each
coming from the coworkers), and the other two con- analysis. Thus, more feminine people appear to react
stituted congruent situations (the level of stressor and more strongly to social support than less feminine or

Table 1
Correlations Among All Variables (N 117)
Variable 1 2 3 4 5 6 7 8
Stressors generated by supervisor
1. Ambiguity
2. Workload .31**
Individual difference
3. Gender role .10 .17
Socially supportive communications
4. With supervisor .31** .15 .10
5. With coworkers .07 .35** .02 .33*
Outcomes
6. Job dissatisfaction .37** .32** .27** .15 .22*
7. Anxiety .25** .47** .00 .10 .24* .33**
8. Psychological strain .46** .50** .01 .20* .29** .52** .76**
* p .05. ** p .01.
SOCIAL SUPPORT 225

Table 2 account for previously inconsistent results, and we


Moderated Regressions of Outcomes on Workload examined one of each type of characteristic: (situa-
and Support tional) congruence between the source of the stressor
Criterion/ordered predictors R2 change and the social support and (personal) gender role
characteristics.
Job dissatisfaction A prerequisite to such a study is to show that
1. Demographic controls .09**
2. Workload (A) .09** 0.46 stressors emanating from the supervisor could indeed
Supervisor support (B) 0.15 lead to individual strains. This seemed very likely,
3. A B .00 0.28 but previous research has largely ignored the source
Job dissatisfaction (person) of stressors as an issue in occupational
1. Demographic controls .09**
2. Workload (A) .10** 0.54
stress. As expected, the supervisor stressors were
Coworker support (B) 0.36 strongly and positively related to the individual
3. A B .01 0.42 strains, supporting Hypothesis 1. One issue that
Anxiety needs further research concerns potential differences
1. Demographic controls .01 between stressors emanating from different sources.
2. Workload (A) .24** 0.96**
Supervisor support (B) 0.61* We would propose, for example, that stressors com-
3. A B .03* 0.74* ing from the supervisors are likely to be more potent
Anxiety than stressors coming from coworkers or people
1. Demographic controls .01
2. Workload (A) .25** 1.54**
Coworker support (B) 0.82**
3. A B .06** 1.53**
Psychological strain
1. Demographic controls .02 Table 3
2. Workload (A) .27** 0.84** Moderated Regressions of Outcomes on Ambiguity
Supervisor support (B) 0.35 and Support
3. A B .02 0.53
Psychological strain Criterion/ordered predictors R2 change
1. Demographic controls .02
Job dissatisfaction
2. Workload (A) .28** 1.18**
1. Demographic controls .09**
Coworker support (B) 0.62*
2. Ambiguity (A) .12** 0.03
3. A B .03* 1.01*
Supervisor support (B) 0.32
Note. Demographic controls were gender and age; betas 3. A B .02 0.39
are from final step (3). Job dissatisfaction
p .10. * p .05. ** p .01. 1. Demographic controls .09**
2. Ambiguity (A) .17** 0.35
Coworker support (B) 0.21
3. A B .00 0.01
Anxiety
1. Demographic controls .01
more masculine people do, supporting Hypothesis 3. 2. Ambiguity (A) .07** 0.50*
In additional analyses, separate masculine and femi- Supervisor support (B) 0.22
nine subscales were formed and were tested for their 3. A B .01* 0.28*
interactions with supervisor and coworker social sup- Anxiety
1. Demographic controls .01
port predicting the three outcomes (with age and sex 2. Ambiguity (A) .12** 0.84
as covariates). None of these interactions were Coworker support (B) 0.48*
significant. 3. A B .01 0.66
Psychological strain
1. Demographic controls .02
Discussion 2. Ambiguity (A) .23** 0.37
Supervisor support (B) 0.13
Although previous research frequently found so- 3. A B .00 0.13
cial support to be related to employee strains via Psychological strain
main effects or interactions with stressors, results for 1. Demographic controls .02
2. Ambiguity (A) .29** 0.67
main effects are often weak, and details of the inter- Coworker support (B) 0.35
actions are often puzzling. The present study sheds 3. A B .00 0.23
light on the puzzle by examining two explanations Note. Demographic controls were gender and age; betas
proposed to explain the inconsistencies. Variations in are from final step (3).
both situational and personal characteristics might p .10. * p .05. ** p .01.
226 BEEHR, FARMER, GLAZER, GUDANOWSKI, AND NAIR

Figure 1. Interactions of supervisor social support and stressors emanating from the
supervisor.

lower in the organizational hierarchy. Because people would lead to one type of interaction (reverse buff-
above us in the hierarchy have more formal power ering), whereas source incongruence would lead to
over us than people lower in the hierarchy, demand another type of interaction (buffering). That is, sup-
characteristics from their messages might have more portive communications from the supervisor would
powerful effects on us in terms of strains. Little make the presumed causal effect of stressors (also
research attention has been given to the effects of emanating from the supervisor) on strains stronger
different sources of potentially stressful demands. rather than weaker, because of the congruence be-
Although it was beyond the scope of this study to tween stressor and support sources. The supervisors
examine this, it would be a timely issue for future presence would make the stressor more salient, even
research. though he or she might be trying to be supportive. In
It was also hypothesized (Hypotheses 2a and 2b), terms of cognitive dissonance, perceiving the oppo-
based on previous suggestions (e.g., Blau, 1981; site (stressing and supporting) behaviors by the su-
Kaufmann & Beehr, 1986), that congruence between pervisor would create unpleasant tension, thus in-
the source of stressor and the source of social support creasing the strain rather than decreasing it. This
SOCIAL SUPPORT 227

Figure 2. Interactions of coworker social support and stressors emanating from the
supervisor.

study attempted to explain the previous puzzling re- be the best person to provide social support, precisely
search results in which sometimes buffering effects because that person is particularly relevant to the
and sometimes reverse buffering effects of social stressful situation. That is, source congruence might
support have been found, but the hypothesis was not have led to favorable (buffering) effects rather than to
supported. There were few significant interactions, unfavorable effects, because the source of the stressor
most involving the strain anxiety and the stressor (the supervisor in this study) might be in the best
workload. position to understand the stressor and to ease the
All interaction effects in the study, in both congru- persons reactions to it. In addition to the interaction
ent and incongruent situations, were buffering ef- effects, the negative correlation between supervisor
fects. Social support tended to make the relationship support and ambiguity suggests that supervisor com-
between stressors and strains weaker regardless of munication might have a main effect, lowering the
the congruence between the source of support and the stressor ambiguity. This is consistent with the notion
source of stressors. An alternative to the source con- that supervisor communication might be useful for
gruence hypothesis in the present study is based on reducing supervisor-induced stress.
the proposition that the source of the stressor might These results regarding congruence lead us to
228 BEEHR, FARMER, GLAZER, GUDANOWSKI, AND NAIR

Table 4 (Viswesvaran et al., 1999). Future research should


Moderated Regressions of Outcomes on Gender examine and measure alternative forms of congru-
Role and Support ence between occupational stressors and social sup-
Criterion/ordered predictors R2 change port. Issues include the nature of the interaction (e.g.,
how logical would it seem to the receiver of support
Job dissatisfaction for someone who causes stress also to offer support),
1. Demographic controls .09**
2. Gender role (A) .05 0.70* the closeness of the relationship (e.g., even though
Supervisor support (B) 1.27 the supervisor is causing the stressor, he or she has
3. A B .03* 1.57* been a friend in the past), and the degree to which the
Job dissatisfaction stressed employee attributes causal blame for the
1. Demographic controls .09**
2. Gender role (A) .08** 0.42
stressor to the supportive person (e.g., the supervisor
Coworker support (B) 0.58 had no choice vs. the supervisor increased my stres-
3. A B .00 0.42 sors of his or her own free will).
Anxiety We also were interested in the potential interac-
1. Demographic controls .01 tions of social support and individual differences
2. Gender role (A) .01 0.32
Supervisor support (B) 0.54 variables in predicting strains. It has often been con-
3. A B .01 0.71 cluded that social support has direct relations to
Anxiety strains (e.g., Beehr, 1995; House, 1981; Vaux, 1988),
1. Demographic controls .01 but the empirical relationships have been somewhat
2. Gender role (A) .06* 0.04
Coworker support (B) 0.18
weak (Beehr, 1995; Viswesvaran, et al., 1999). One
3. A B .00 0.07 reason for these weak relationships could be that
Psychological strain there are individual difference moderators of the re-
1. Demographic controls .02 lationship; that is, the relationship is stronger for
2. Gender role (A) .03 0.53 some people and weaker or nonexistent for other
Supervisor support (B) 1.18
3. A B .03 1.50 people, perhaps because some people value social
Psychological strain support more than others (ODriscoll & Dewe, 2001).
1. Demographic controls .02 The present study only looked at one such individual
2. Gender role (A) .08** 0.09 difference (gender role), but future research on others
Coworker support (B) 0.13
3. A B .00 0.19
seems promising.
Traditional stereotypes consider women to be
Note. Demographic controls were gender and age; betas more social in nature than men, and there is probably
are from final step (3).
p .10. * p .05. ** p .01. some truth to this. Women have often been reported
to provide more advice to and do more favors for
people experiencing problems than men (e.g., a re-
view by Eagley & Crowley, 1986). They would also
benefit more when on the receiving end of such social
question the nature of the concept of congruence. We support (Perrewe & Carlson, 2002). Rather than gen-
hypothesized that source congruence (i.e., sources of der itself, however, this effect is more likely due to
stressors and supports are the same) would be more assuming the so-called feminine gender role, which
harmful than helpful. This did not happen, however, we focused on in the present study. It was expected
and the nature of congruence might need revisiting. that peoples gender roles would be related to their
Indeed, forms of congruence other than the type in social nature, and that people with more feminine
the present study have also been proposed as expla- gender role characteristics would react more posi-
nations for inconsistent interaction effects, and these tively to social support than people with less femi-
should be the focus of future research. Specificity nine or more masculine role characteristics would. In
effects were proposed by Cohen and Wills (1985) the present data, there was some evidence that this
regarding instrumental support having a buffering was true. Employees with more feminine gender
effect when the supportive behavior is congruent roles appear to react more positively to social sup-
with or specific to the stressor, and Frese (1999) port. This could be one of the explanations for pre-
found results that were somewhat consistent with vious weak results regarding the relationship between
this. Researchers have had great difficulty trying to social support and strains in general (Viswesvaran et
predict in advance when a specific type of social al., 1999). For some people, the relationship might be
support will result in buffering and when it will not, substantially stronger. This helps to explain the enig-
although explanations are often given after the fact matic discrepancy between the common belief that
SOCIAL SUPPORT 229

Figure 3. Interactions of gender role and supervisor social support.

social support helps relieve strains and the relatively limiting the power to find significant interactions.
poorer and more inconsistent record of finding strong This is probably related to the fact that only 5 of the
empirical evidence for this relationship. In our every- 18 total interactions tested were significant at the
day lives, we may indeed observe positive effects for traditional p .05 level. We relaxed the significance
social support, but these nonsystematic observations criteria to consider results that were p .10, and this
may be biased toward the instances in which it works resulted in consideration of any interaction for which
well. One such instance is apparently when the re- the R2 change was greater than .02. Thus, the best
ceiver of the social support is relatively more femi- estimate is that only small interactions were not de-
nine in his or her gender role. tected due to sample size. Second, all data were
obtained by means of a single survey at one time,
Limitations possibly inflating some relationships artificially and
making strong causal inference impossible. High cor-
There were several limitations in the study. First, relations between the stressor and the moderator can
the sample size was relatively small (N 117), thus also lead to difficulty finding interactions. Based on
230 BEEHR, FARMER, GLAZER, GUDANOWSKI, AND NAIR

the correlations in Table 1, this would primarily be a to use theoretical or rational approaches for choosing
problem for interactions of ambiguity with supervisor potential moderators to examine.
support and of workload with coworker support, in
which there were moderate correlations between the
stressor and the support (.31 and .35, respectively). References
Multiple regressions in Tables 2 and 3 show, how-
Banks, M. H., Clegg, C. W., Jackson, P. R., Kemp, N. J.,
ever, that there were actually more rather than fewer Stafford, E. M., & Wall, T. D. (1980). The use of the
significant interactions with these combinations than General Health Questionnaire as an indicator of mental
with other combinations of stressors and supports. health in occupational studies. Journal of Occupational
Thus it does not appear that this problem accounted Psychology, 53, 187194.
for the pattern of significant interactions. Third, for Beehr, T. A. (1995). Psychological stress in the workplace.
London: Routledge.
the significant interactions, most of the changes in Beehr, T. A., & Glazer, S. (2001). A cultural perspective of
slopes were not large, indicating that the moderating social support in relation to occupational stress. In P. L.
effects might lead only to small differences in out- Perrewe & D. C. Ganster (Eds.), Research in occupa-
comes for the employees. Fourth, regarding general- tional stress and well being: Exploring theoretical mech-
anisms and perspectives (Vol. 1, pp. 97142). New York:
ization of the results to other samples, it should be JAI Press, Elsevier Science.
noted that the means of the stressor and strain vari- Beehr, T. A., Jex, S. M., Stacy, B. A., & Murray, M. (2000).
ables were near the low end of the scales. This raises Work stressors and coworker support as predictors of
the issue of whether the results would have been the individual strain and job performance. Journal of Orga-
same for groups of employees experiencing more nizational Behavior, 21, 391 405
Beehr, T. A., King, L. A., & King, D. W. (1990). Social
stress. A fifth limitation concerns the nature of social support and occupational stress: Talking to supervisors.
support: Are the measures of communication actually Journal of Vocational Behavior, 36, 61 81.
measures of social support? They were developed as Beehr, T. A., Walsh, J. T., & Taber T. D. (1976). Relation-
new and different operationalizations of social sup- ship of stress to individually and organizationally valued
states: Higher order need strength as a moderator. Jour-
port because of the weaknesses and inconsistent re- nal of Applied Psychology, 61, 41 47.
sults using other social support measures (Beehr et Bem, S. L. (1974). The measurement of psychological an-
al., 1990), and all subsequent research has considered drogyny. Journal of Consulting and Clinical Psychology,
them to be measures of social support. They are very 42, 155162.
specific indices, however, measuring contents of Billings, A. G., & Moos, R. H. (1982). Work stress and the
stress-buffering roles of work and family resources.
communication Journal of Occupational Behaviour, 3, 215232.
Blau, G. (1981). An empirical investigation of job stress,
social support, service length and job strain. Organiza-
tional Behavior and Human Performance, 27, 279 302.
Future Research Cammann, C. C., Fichman, M., Jenkins, G. D., Jr., & Klesh,
J. (1979). The Michigan Organizational Assessment
Future research should continue to examine addi- Questionnaire. Ann Arbor: University of Michigan.
tional ways in which social support plays a role in Caplan, R. D., Cobb, S., French, J. R. P., Jr., Harrison,
occupational stress. Past research, theory, and prac- R. V., & Pinneau, S. R., Jr. (1975). Job demands and
tice have focused on social support as a key variable worker health: Main effects and occupational differ-
ences. Washington, DC: U.S. Government Printing
for mitigating the effects of stress, but the amount of Office.
variance in strains that social support can explain Chen, P. Y., Popovich, P. M., & Kogan, M. (1999). Lets
varies a great deal. As ODriscoll and Dewe (2001) talk: Patterns and correlates of social support among
have noted, the challenge is to determine and predict temporary employees. Journal of Occupational Health
Psychology 4, 55 62.
how the strength of this empirical relationship varies Cohen, S., & Wills, T. A. (1985). Stress, social support, and
in systematic ways; that is, theoretically, what are the the buffering hypothesis. Psychological Bulletin, 98,
likely moderator variables? The answer to this ques- 310 357.
tion will determine the types of social support inter- Eagley, A. H., & Crowley, M. (1986). Gender and helping
ventions that can alleviate the effects of job stressors behavior: A meta-analytic review of the social psycho-
logical literature. Psychological Bulletin, 100, 283308.
on strains such as anxiety, depression, burnout, and Fenlason, K. J., & Beehr, T. A. (1994). Social support and
the attrition associated with them. Although social occupational stress: effects of talking to others. Journal
support is thought to be a powerful variable in occu- of Organizational Behavior, 15, 157175.
pational stress episodes, the aims are to mold and Festinger, L. (1957). A theory of cognitive dissonance.
Stanford, CA: Stanford University Press.
target social support applications in more successful Frese, M. (1999). Social support as a moderator of the
ways. As Viswesvaran et al. (1999) noted, and as relationship between work stressors and psychological
exhibited in the present study, future research needs dysfunctioning: A longitudinal study with objective mea-
SOCIAL SUPPORT 231

sures. Journal of Occupational Health Psychology, 4, McIntosh, N. J. (1991). Identification and investigation of
179 192. properties of social support. Journal of Organizational
Ganster, D. C., Fusilier, M. R., & Mayes, B. T. (1986). Role Behavior, 12, 201217.
of social support in the experience of stress at work. ODriscoll, M. P., & Dewe, P. J. (2001). Mediators and
Journal of Applied Psychology, 71, 102110. moderators of stressorstrain linkages. In P. L. Perrewe
Glaser, D. N., Tatum, C. B., Nebeker, D. M., Sorenson, & D. C. Ganster (Eds.), Research in occupational stress
R. C., & Aiello, J. R. (1999). Workload and social and well being: Exploring theoretical mechanisms and
support: Effects on performance and stress. Human Per- perspectives (Vol. 1, pp. 257 287). New York: JAI
formance, 12, 155176. Press, Elsevier Science.
Goldberg, D. P. (1972). The detection of psychiatric illness Perrewe , P. L., & Carlson, D. S. (2002). Do men and
by questionnaire: A technique for the identification and women benefit from social support equally? Results from
assessment of non-psychotic psychiatric illness. London: a field examination within the work and family context.
Oxford University Press. In D. L. Nelson & R. J. Burke (Eds.), Gender, work
Greenglass, E. R., & Burke, R. J. (1988). Work and family stress, and health: Current research issues (pp. 101
precursors of burnout in teachers: Sex differences. Sex 114). Washington, DC: American Psychological
Roles, 18, 215229. Association.
Harmon-Jones, E. (1999). Toward an understanding of the Spector, P. E., Dwyer, D. J., & Jex, S. M. (1988). Relation
motivation underlying dissonance: Is the production of of job stressors to affective, health, and performance
aversive consequences necessary to produce dissonance? outcomes: A comparison of multiple data sources. Jour-
In E. Harmon-Jones & J. Mills (Eds.), Cognitive disso- nal of Applied Psychology, 73, 1119.
nance: Progress on a pivotal theory in social psychology Spielberger, C. (1979). Preliminary manual for the State
(pp. 7199). Washington, DC: American Psychological Trait Personality Inventory. Unpublished manuscript,
Association. University of South Florida.
House, J. S. (1981). Work stress and social support. Read- Stephens, C., & Long, N. (2000). Communication with
ing, MA: Addison-Wesley. police supervisors and peers as a buffer of work-related
Kahn, R. L., & Byosiere, P. (1992). Stress in organizations. traumatic stress. Journal of Organizational Behavior, 21,
In M. D. Dunnette & L. M. Hough (Eds.), Handbook of 407 424.
industrial and organizational psychology (Vol. 3, 2nd Vaux, A. (1988). Social support: Theory, research, and
ed., pp. 571 650). Palo Alto, CA: Consulting Psycholo- intervention. New York: Praeger Press.
gists Press. Vermeulen, M., & Mustard, C. (2000). Gender differences
Kaufmann, G. M., & Beehr, T. A. (1986). Interactions in job strain, social support at work, and psychological
between job stressors and social support: Some counter- distress. Journal of Occupational Health Psychology, 5,
intuitive results. Journal of Applied Psychology, 71, 428 440.
522526. Viswesvaran, C., Sanchez, J. I., & Fisher, J. (1999). The
King, L. A., King, D. W., Fairbank, J. A., Keane, T. M., & role of social support in the process of work stress: A
Adams, G. A. (1998). Resilience-recovery factors in meta-analysis. Journal of Vocational Behavior, 54,
post-traumatic stress disorder among female and male 314 334.
Vietnam veterans: Hardiness, postwar social support, and Wade, T. D., & Kendler, K. S. (2000). Absence of interac-
additional stressful life events. Journal of Personality tions between social support and stressful life events in
and Social Psychology, 74, 420 434. the prediction of major depression and depressive symp-
La Rocco, J. M., & Jones, A. P. (1978). Co-worker and tomatology in women. Psychological Medicine, 30,
leader support as moderators of stressstrain relation- 965974.
ships in work situations. Journal of Applied Psychology,
63, 629 634.
Lim, V. K. G. (1997). Moderating effects of work-based
Received August 23, 2002
support on the relationship between job insecurity and its Revision received January 17, 2003
consequences. Work & Stress, 11, 251266. Accepted March 9, 2003 y

Potrebbero piacerti anche