Journal of Personality and Social Psychology Copyright 1987 by the American Psychological Association, Inc.

1987, Vol. 53. No. 6, 1137-1145 0022-3514/87/S00.75

Determinants of Perceived Social Support: Interpersonal Transactions,

Personal Outlook, and Transient Affective States

Amiram Vinokur Yaacov Schul

Institute for Social Research Hebrew University of Jerusalem
University of Michigan Jerusalem, Israel

Robert D. Caplan
Institute for Social Research
University of Michigan

During the last decade, the study of social support has relied heavily on recipients' reports of per-
ceived support. However, such reports of support may reflect not only actual supportive interpersonal
transactions but also the recipients' own personal and perceptual dispositions and comparatively
transient mood states. This study examines these factors' independent effects on perceived support.
Interpersonal transactions of social support were assessed in a longitudinal panel design using reports
from 486 adult male respondents and their significant others (mostly wives). Using structural model-
ing techniques (LISREL vi), an unfolding series of statistical analyses were performed. The results
provide strong support for a model in which recipients' perception of support is determined signifi-
cantly and strongly by actual interpersonal transactions as reported by significant others, moderately
by the recipients' negative outlook bias, and weakly by their anxiety and depression as indicators of
their poor mental health. The implications of the results for future research are discussed.

The last 15 years have witnessed a spectacular growth of re- have relied on measures of the availability of a social network,
search on the effects of social support on health and well-being. on measures of frequency of social contacts (e.g., Berkman &
Hundreds of research investigations have been conducted on Syme, 1979; House, Robbins, & Metzner, 1982), or more di-
the subject during this period. Yet, many fundamental issues rectly, on measures of the perceived availability of support (e.g.,
remain unresolved. Typically, researchers have focused on ei- Cohen, Mermelstein, Kamarck, & Hoberman, 1985; I. G. Sara-
ther the social structural aspects of social support by examining son, Levine, Basham, & Sarason, 1983). Studies that have fo-
social contacts and networks, or they have focused on the psy- cused on support as an interpersonal transaction have exam-
chological aspects by examining perceived obtained support or ined the provision of various kinds of support via behavioral
its perceived availability. Even when the psychological aspects acts as perceived by the recipients (Caplan, Cobb, French, Har-
of social support have been examined, the role of personality rison, & Pinneau, 1975; House, 1981; LaRocco, House, &
has, with rare exceptions (Kobasa & Puccetti, 1983; Lefcourt, French, 1980; Procidano & Heller, 1983). (For elaborate re-
Martin, & Saleh, 1984; B. R. Sarason, Sarason, Hacker, & Bas- views of social support measures and concepts, see Barrera,
ham, 1985; I. G. Sarason, Sarason, & Shearin, 1986), largely 1986; House & Kahn, 1985; and B. R. Sarason, Shearin, Pierce,
been ignored. The social psychological aspects of this phenome- & Sarason, 1987). In either case, the attempt to identify sup-
non—the study of social support as an interpersonal transac- portive transactions has relied on measures of perceived sup-
tion that involves both a provider of support and a recipient— port or potential available support. There has been no informa-
is nearly absent. Our goal in this article is to fill this gap by tion on the extent to which perceived obtained support reflects
examining simultaneously the transactional and cognitive-emo- actual supportive behaviors.
tional determinants of perceived obtained social support. Without such information it is possible that measures of per-
The recent literature on social support addresses two concep- ceived obtained support, and in particular, measures of per-
tually distinct definitions of the concept: social support as the ceived availability of support, are largely a function of personal-
perceived availability of support resources and social support as ity dispositions or processes that consistently bias the percep-
an interpersonal transaction (e.g., Wethington & Kessler, 1986). tion of obtained or potentially available support. This
Studies that have focused on availability of support resources possibility is suggested by correlations among measures of per-
ceived support from different sources and across time periods.
Reanalyses of data from the study reported here revealed that
This research was supported by National Institute of Mental Health the correlations among measures of perceived social support
Grants MH34586 and MH39675 and by the Michigan Prevention Re-
from spouse, close friend, co-worker, and supervisor ranged
search Center funded by National Institute of Mental Health Grant
from . 17 to .59 and averaged .34. The stability of perceived so-
MH38330. We wish to thank Laura KJem Tor help and advice regarding
cial support from all the aforementioned sources across time
data analysis.
Correspondence concerning this article should be addressed to periods from 4 to 8 months ranged from .49 to .69 and averaged
Amiram Vinokur, Institute for Social Research, The University of .60. I. G. Sarason et al. (1986), using measures of perceived
Michigan, Ann Arbor, Michigan 48106-1248. availability of support, found greater stability coefficients. Such

1137
1138 A. VINOKUR, Y. SCHUL, AND R. CAPLAN

correlations suggest (but do not demonstrate) that the personal- that include not only a measure of obtained support as reported
ity of the respondent produces similar and consistent percep- by the recipient but also a measure of provided support as re-
tions and interpretations of supportive behaviors and their ported by the provider. Our purpose is to focus on and examine
availability. the perception of obtained social support as a function of the
One might also hypothesize that the stability of social sup- following three variables: (a) actual provision of social support
port across settings reflects the consistency of the environments from interpersonal transaction as reported by those who pro-
themselves. This could occur because people tend to elicit, or vide it; (b) stable and general personal disposition to perceive
mobilize, the same supportive behaviors across settings or self- negatively oneself and one's social transactions and condition;
select into environments that are compatible with their per- and (c) affective states such as anxiety and depression that are
sonal needs (e.g., Emmons, Diener, Larsen, 1986), and that often used as the main indicators of poor mental health.
provide a consistent level of support. These "environmental" There is ground to hypothesize that each of the aforemen-
explanations depend on dispositions of the recipient of social tioned factors contributes significantly to perceived social sup-
support. Accordingly, they are compatible with a general hy- port. First, the provision of support is expected to have at least
pothesis regarding the role of personality as a plausible determi- minimal impact on perceived support, particularly in enduring
nant of perceived obtained social support. social relationships. In the study that follows, the providers and
Other studies have suggested that personality and early social- the recipients of social support were mostly spouses or close
ization predict to social support. Kobasa and Puccetti (1983) friends in such enduring relationships.
suggested that psychological hardiness influences the degree to Second, it is reasonable to expect that stable cognitive or at-
which social support is successfully mobilized to moderate the tributional styles affect how social support offered by others is
impact of stressful life events on well-being. In their study, psy- being perceived, interpreted, and experienced. People who are
chological hardiness and support from the boss were signifi- disposed to view themselves and their personal experiences in
cantly correlated (r = .29, p < .005). Sandier and Lakey (1982) a negative light are thus hypothesized to misperceive or under-
and Lefcourt et al. (1984) demonstrated that people with an rate the support provided to them. Watson and Clark (1984)
internal locus of control benefit from social support more than have recently suggested that such a disposition constitutes a sta-
those who have a more external orientation. Finally, I. G. Sara- ble and pervasive trait that is latent across diverse personality
son et al. (1986) have suggested parental bonding as a develop- scales. This disposition characterizes people who are more
mental origin to social support and its stability. likely to experience negative affective states and "tend differen-
The issue of whether personality variables are the key deter- tially to dwell on the negative side of themselves and the world"
minants of either the mobilization of support or its perception
(p. 465). This disposition will be referred to in this study as a
is critical for understanding the links between social support
generalized negative outlook. (For an integrative review of this
with health and well-being. As already suggested by Cohen and
disposition, see Watson & Clark, 1984.)
Syme (1985), conclusive evidence for the causal effects of actual
Third, studies of social support typically assume on the basis
supportive transactions, or for the operation of stable biasing
of correlational evidence that social support has a main effect
tendencies in the perception of social support, could be ob-
on mental health and well-being. Nevertheless, it is equally
tained from experimental studies in which support is manipu-
plausible to hypothesize that poor mental health, and in partic-
lated. To date, the experimental literature on this issue is still
ular its main indicators (i.e., depression and anxiety), is a deter-
limited because of the narrow range of social support manipu-
minant of perceived low social support. The literature on affect
lations in laboratory settings (e.g., I. G. Sarason & Sarason,
and memory suggests that mood states influence the likelihood
1986). Until the causal roles played by social support versus
of recall of mood congruent events (see the extensive review
those played by personality are made explicit, the enormous lit-
by Blaney, 1986; Lewinsohn & Hoberman, 1982). Accordingly,
erature on the main and buffering effects of social support (Co-
hen & Wills, 1985; Kessler & McLeod, 1985) will be open to whereas social support may have a beneficial effect on well-be-
equivocal interpretations (Cohen & Syme, 1985; pp. 16-18). ing, we hypothesize that its perception is influenced by concur-
Another methodological approach, besides the experimental rent affective states such as depression and anxiety. The experi-
one, for identifying causal effects of actual supportive transac- ence of these and other negative affective states will invoke more
tions and of stable perceptual dispositions is to use valid exter- memories of negative than positive instances of social interac-
nal measures of provided support as well as personality mea- tions and social support.
sures of relevant perceptual dispositions. External measures To examine these hypotheses, longitudinal data collected
can be obtained from observers or from the provider of support. from a sample of men and their significant others (mostly wives)
In the investigation reported here, we used this approach by were analyzed using the structural modeling techniques of LIS-
securing measures of support from both the providers and the REL vi (Joreskog & Sorbom, 1984). These techniques allowed
recipients of support. Recognize, however, that measures of us to model a latent variable, described as generalized negative
provided support may also be affected by the perceptual biases outlook, based on repeated measures of resentment and self-
of the provider of support. Nevertheless, the degree of the agree- esteem and on a measure of perceived parental support during
ment between measures of provided support with measures of adolescence. An additional advantage of the application of
obtained support from the two sources offers convergent valid- structural modeling is that the estimates of the relative strength
ity for the occurrence of actual interpersonal supportive trans- of influence of the independent variables are more reliable than
actions. in path analysis. Structural modeling makes it possible to take
In this article we address the question regarding the determi- into account the unreliability associated with the various mea-
nants of perceived obtained social support by examining data sures.
 DETERMINANTS OF PERCEIVED SOCIAL SUPPORT 1139

Method ers included six questions on the provision of socially supportive behav-
iors to the focal respondent (e.g., showing care and concern, providing
Overview of the Sample and Design useful information). These questions were the same as those included
in the interview with the focal respondent (described later) except that
This study is based on data collected as part of a larger investigation they referred to providing rather than receiving support. The measure
on stress, work, and unemployment among Vietnam veterans and non- of significant-other-provided support had alpha coefficients ranging
veterans (for a complete report, see Vinokur, Caplan, & Williams, in from .81 to .85 for the three data collection waves.
press). The analyses reported here were derived from data collected at Obtained social support. The measure for this construct assessed per-
three time periods from a panel of 486 male respondents and their sig- ceived supportive behaviors of the significant other. It was based on a 6-
nificant others (wives and close friends). Of the male respondents, 297 item index, with coefficient alphas ranging from .85 to .87 for the three
had recently lost their jobs and were unemployed at the first wave of waves of data collection. To measure social support, we asked the re-
data collection. The other 189 respondents were employed. The first spondents to indicate on 5-point scales, which ranged from not at all ( I )
wave of data was collected in February and March 1982. The second to a great deal (5), how much the significant other "provides you with
and third waves took place 4 months and 12 months, respectively, fol- encouragement" and with "useful information," "says things that raise
lowing the first one. your self-confidence," "listens to you when you need to talk," "shows
that he/she cares about you as a person," and "understands the way you
Respondents and Procedure think and feel about things." The social support items were chosen to
represent three of the four functions of social support suggested by
The sample of respondents was drawn to represent unemployed Viet- House (1981). These functions include emotional, appraisal, and infor-
nam veterans, era veterans (in the military during the Vietnam war but mational support. Instrumental support was not assessed by these
stationed elsewhere), nonveterans of the same generation, and their em- items.
ployed counterparts. The unemployed respondents were recruited from Poor menial health. In this study we measured poor mental health
nine state unemployment offices in southeast Michigan. The employed symptomatology using Anxiety and Depression subscales based largely
respondents were nominated by their unemployed counterparts to pro- on the Hopkins Symptom Checklist (Derogatis, Lipman, Rickels, Uh-
duce demographically matched groups. This intention was met success- lenhuth, & Covi, 1974). The measures of anxiety and depression in-
fully. Eligible unemployed respondents were men between 24 and 41 cluded 7 items each, with alphas ranging from .85 to .87 across the three
years of age who had lost their jobs within the last 4 weeks and were time waves. The questions in these scales required the respondent to
working for pay fewer than 20 hr a week or were not working at all. "tell how often you have experienced each of the following in the last
The specific sampling design was intended to serve the purpose of the two weeks" The respondent was then read symptoms such as "nervous-
original, larger investigation (Vinokur et al., in press) and does not con- ness or shakiness inside," "worrying about things," "feeling blue," and
cern the present study. "feeling depressed" and was asked to rate the symptom's frequency on
At each interview, the focal respondent selected a significant other a 5-point scale, which ranged from not at all (1) to very often (5).
who was asked to complete a self-administered questionnaire. The sig-
nificant other was denned as someone who knew the respondent well
Indicator Measures of Generalized Negative Outlook
and who saw the respondent at least once a week. Seventy-one percent
of the significant others were the wives of the respondents, 16% were Three indicators were chosen for assessing the stable personality dis-
girlfriends or close friends, and 13% were others, such as roommates, position referred to here as generalized negative outlook. Our choices
relatives, and neighbors. were based on the indicators' broad representation of the concept as
Data collection. Data from the respondents were collected through defined here and by Watson and Clark (1984). The measures covered the
personal interviews. The interviews took place at the respondent's loca- respondent's outlook on his past relationship with his parents (Parental
tion of choice, usually his home, and they lasted for about 1 hr. Upon Support scale) and on his present view of his ordeal in life and with
completion of each interview, the respondent was paid $5 for his partici- people (Resentment scale) and his view of himself (Self-Esteem scale).
pation. Data from the significant others were collected using self-admin- The latent variable represented the common variance among these indi-
istered questionnaires that took approximately 20 min to complete. cators—variance that was assumed to transcend the particular content
The significant others were paid $2 for their participation at each time of each indication and to represent a generalized perceptual disposition.
wave. Parental support during childhood and adolescence was assessed by
Response rate. Of those men who were found eligible, 84% were re- indexes derived from the work of Gold and his associates (Gold &
cruited to the study as focal respondents and were interviewed in Wave Mann, 1972; Gold & Reimer, 1975). A Parental Supportiveness scale
1. Continued participation in the study was even higher. The response was constructed from two 8-item indexes regarding felt closeness to,
rate for Wave 2 was 95%, with 462 of the original 486 focal respondents identification with, and support received from father and mother (or
participating in the study. At Wave 3,96%, or 442, of the Wave 2 respon- guardians) during one's teenage years. The items included ratings of
dents were interviewed. Similarly high response rates were obtained items such as "How much did your (father/mother/guardian) accept
with the significant others. In each time wave, data were collected from and understand you as a person?" "How close did you feel to your (fa-
more than 90% of the respondents' significant others. ther/mother/guardian)?" "How much did he/she try to help you when
you were scared or upset?" and "How much did he/she show you affec-
tion?" The rating scale ranged from not at all (1) to a great deal (5}. The
Measures
Parental Supportiveness scale for father and for mother had an alpha
Most of the measures in this study were based on instruments used coefficient of .85.
in earlier investigations on stress and unemployment and on social sup- Resentment was measured using a 5-item index developed by Blu-
port. In most instances, we used multi-item indexes to maximize the menthal. Kahn, Andrews, and Head (1972). The measure had alpha
internal consistency of the measures. The Cronbach alpha coefficients coefficients ranging from .80 to .87 across our three waves of data collec-
of our measures ranged from .80 to .87. We constructed the respective tion. The items required the respondent to indicate on 5-point scales the
index measures by averaging the ratings to the items that made up extent of agreement or disagreement with various statements reflecting
the measures. (Copies of the measures can be obtained from Amiram resentment and paranoia. The respondents were asked to "tell how
Vinokur.) much you agree or disagree with. . ." the statements "I don't seem to
Provided social support. The questionnaire from the significant oth- get what is coming to me," "when I look back on what happens to me I
1140 A. VINOKUR, Y. SCHUL, AND R. CAPLAN

feel resentful," "I feel I get a raw deal out of life," "people ask too much fore linked to represent correlated errors that were set to be
of me," and "other people always seem to get the breaks." equal.
The self-esteem measure was an index based on seven items with co- The estimation of this initial model produced a chi-square of
efficient alphas ranging from .82 to .83 across the data collection waves.
1774 with 710 degrees of freedom. This figure, which is 2.5
The items asked the respondent to indicate the extent to which he was
times the degrees of freedom, represents a fairly good fit be-
satisfied with himself, felt confident, compared himself favorably with
tween the model and the data.
the average man, saw himself as a failure, was able to handle problems
in his life, and was happy with his accomplishments. The estimated model produced high betas between the con-
Age, education, and income were assessed using standard survey re- structs at Time 1 and Time 2. That is, each of the endogenous
search questions that will not be detailed here. variables at Time 1 was an excellent predictor of its level at the
subsequent data collection at Time 2. These betas varied only
slightly from .66 to .71.
Results

Overview of the Analyses Extended Structural Model

We viewed the model in Figure 1 as incomplete because it

On the basis of our hypotheses, we generated first an initial
implied that the stabilities of the endogenous variables were due
simplified longitudinal model with generalized negative out-
to the effects of their antecedent levels on the consequent levels.
look, resentment, self-esteem, and parental support as the exog-
As an alternative, we generated an extended structural model
enous conceptual variables and perceived obtained social sup-
in which stable personality dispositions and social relationships
port, perceived provided support, and poor mental health as the
between the significant other and the focal respondent would
endogenous conceptual variables. All the endogenous variables
account for the stability of the respondent's perception of ob-
at Time 1 were modeled as predictors of the same measures at
tained support. In addition, we included the respondent's emo-
Time 2, respectively. In the second phase of the analysis, a series
tional state (i.e., mood) as an additional determinant of the ob-
of increasingly more elaborate causal models were constructed,
tained support based on the previously cited literature about
and their goodness of fit was tested against the initial model
the effects of mood on cognition. (We also considered a model
and, then, against the preceding less elaborate models in the
in which obtained support influences mood, the results of
series. Finally, partial replication of the best fitting model was
which will be described later.)
provided by testing it again to predict Time 3 social support
We hypothesized that the recipient's generalized negative
from the corresponding Time 2 variables.
outlook would have the strongest influence on obtained social
support and poor mental health and the smallest influence on
Initial Structural Model the provider's sent social support. This latter link was hypothe-
sized as being due to the provider's possible aversion to helping
The LISREL vi procedure (Joreskog & Sorbom, 1984) was people with a negative disposition (e.g., Coyne, 1976a, 1976b).
used to estimate the parameters of the initial structural model. As a stable personality disposition, the generalized negative out-
The results of this analysis are presented in Figure 1. look was also hypothesized to account for the stability observed
The initial structural model consisted of two unconnected in the level of obtained social support and poor mental health
parts. The first part included the following four exogenous con- between Time 1 and Time 2. If this hypothesis is correct, the
structs: (a) resentment, (b) self-esteem, (c) parental support, stability (i.e., betas) of the obtained social support and of poor
and (d) generalized negative outlook. The first two constructs mental health between Time 1 and Time 2 should decrease
were indicated by the corresponding measures from Time 1, 2, when the influence of the generalized negative outlook on these
and 3 data collections, and their loadings were set to be equal. variables is modeled in our analyses.
The self-esteem construct was estimated using subindexes of
positively and negatively keyed items. The parental support Linking Provided Social Support and Poor Mental
construct was measured at Time 3 only and was indicated by the
Health to Obtained Support
respective subscales regarding support from mother and from
father (or guardian). The generalized negative outlook con- Using the conceptual framework just described, we first
struct was estimated in the structural model using the Resent- tested a series of successive LISREL causal models that included
ment and Self-Esteem scales (with measures from Time 1, 2, systematic variations in the links among the endogenous vari-
and 3 data collections) and perceived support from mother and ables. We started with a model that included a path of influence
from father during childhood and adolescence (measured at from provided social support to obtained social support at each
Time 3 only). time wave. This path represents the extent to which perceived
The second part of the initial structural model included the social support actually reflects interpersonal transactions where
three endogenous constructs: obtained social support, provided support is provided. This model provided a significantly better
social support, and poor mental health. Each construct at Time fit, x2(709, AT = 440) = 1694, p < .001, than the initial one,
1 was modeled to influence its level at Time 2. The indicators X 2 (710,JV=440)= 1774.
for the social support constructs were the questionnaire items; Next, we added to the model the additional paths of influence
those for the mental health construct were the Anxiety and De- from poor mental health to obtained social support at each time
pression subscales. We also assumed that the responses to the wave. This addition resulted in a significantly better fitting
same questions at different times would share some unique vari- model, x2(708, N = 440) = 1619, p < .001, than the previous
ance. The residuals associated with these responses were there- one. A model with additional paths from obtained social sup-
 DETERMINANTS OF PERCEIVED SOCIAL SUPPORT 1141

Time 1 Data Time 2 Data

FR
General
Negative
Outlook
FR
Obtains
Social
Support

Figure 1. Initial model predicting the focal respondent's (FR) obtained social support, the significant other's
(SO) provision of social support, and the PR's poor mental health and generalized negative outlook as
indicated by measures of parental support, self-esteem, and resentment. (The letter r refers to the wave of
data collection. All the beta coefficients were statistically significant beyond the .001 level. The chi square
was 1774 with 710 degrees of freedom.)

port at Time 1 to poor mental health and to provided social tween generalized negative outlook and the endogenous vari-
support at Time 2 neither improved the fit nor demonstrated ables.
significant changes in the coefficients of the previous model. Figure 2 presents the best fitting model produced by these
successive model testings. It contains the paths of influence that
Linking Generalized Negative Outlook and Its Elements were statistically significant beyond the .05 level.
to Social Support and Mental Health The LISREL analysis of this model yielded a chi-square of
1369 with 704 degrees of freedom. This result constitutes sub-
Next, we shifted to a new series of analyses that tested models stantial improvement over the results of the initial structural
with successive additions of paths of influence from the exoge- model with which we started. The ratio of chi-square to degrees
nous variables (e.g., generalized negative outlook) to the endog- of freedom for this model is 1.94, compared with the initial
enous variables (e.g., obtained social support). In this series of model ratio of 2.5. The improvement is statistically significant
analyses the paths of influence among the endogenous variables beyond the .001 level. Furthermore, Hoelter's (1983) critical n
were kept the same as in the best fitting model described earlier. index of goodness of fit was calculated and yielded a value
In the first model tested in this series, generalized negative of 247. A value above 200 is indicative of a model that ade-
outlook was modeled to have a path of influence on each of quately reproduces the observed covariance structure (Hoelter,
the endogenous variables. This model produced a significantly 1983).
better fit x2(705, ff = 440) = 1406, p < .001, than the previous In this extended model, obtained social support at Times 1
best fitting model. Finally, we tested models with additional and 2 was affected in a similar way by the other variables. Pro-
paths of direct influence from the measures of parental support, vided social support had the strongest influence on obtained
self-esteem, and resentment to each of the endogenous vari- social support, with betas of .35 and .30 at Times 1 and 2, re-
ables. Again, these models included all the significant paths be- spectively. Thus there is convergent validation for the role of
1142 A. VINOKUR, Y. SCHUL, AND R. CAPLAN

Time 1 Data Time 2 Data

so
Provides fi4 / Provides
Social -' Social
Support Support
-.07
-.07 '

j.35 .30 .47

FR FR FR
General -.17 Obtains 44 / Obtains
Negative Social ' -H Social
Outlook Support Support

-.15
-.17
.52 .48 ^
FR .64 FR \ .42^
Poor 29 / Poor
Mental -H Mental
.31' Health Health

.28
FR
Resentment

Figure 2, Causal model of the effects of the significant other's (SO) provision of social support, focal respon-
dent's (FR) poor mental health, PR's generalized negative outlook, and resentment upon PR's obtained
social support. (All the beta coefficients were statistically significant beyond the .05 level. The chi square
was 1369 with 704 degrees of freedom.)

interpersonal transactions in influencing perceived obtained so- look as part of the larger disposition to experience aversive emo-
cial support. tional states.
Generalized negative outlook was also found to be a signifi- Finally, poor mental health was also found to have a signifi-
cant determinant of the perception of obtained support at both cant influence on obtained social support, with betas of-.18
time periods {betas of—.17 and —.15, respectively, for Times 1 and -.17 for Times 1 and 2, respectively. As hypothesized, the
and 2). The more negative the outlook, the less the obtained negative affective states of anxiety and depression appear to de-
support. Furthermore, in both time waves the path coefficients crease or suppress the perception of socially supportive acts
between negative outlook and provided support were -.07 from the significant other.
(p < .05). It thus appears that negative outlook had a small, yet As predicted, the effects of provided social support, general-
statistically significant, negative influence on the inclination of ized negative outlook, and poor mental health reduced the
the significant other to provide support. across-time beta stability coefficient of obtained social support
In addition to its influence on obtained social support, gener- from .69 (cf. Figure 1) to .44 (cf. Figure 2). Similarly, the effects
alized negative outlook was found to have a very strong influ- of generalized negative outcome and resentment reduced the
ence on poor mental health, with betas of .52 and .48 for Times across-time beta stability coefficient of poor mental health from
1 and 2, respectively. Similarly, resentment was found to have a .71 {cf. Figure 1) to .29 (cf. Figure 2).
direct influence on poor mental health, with betas of .31 and In the extended model presented in Figure 2, poor mental
.28 for Time 1 and Time 2. The effects of resentment and of health was modeled as a determinant of obtained social sup-
negative outlook on poor mental health were congruent with port. An equally cogent case can be made in favor of modeling
Watson and Clark's (1984) conceptualization of negative out- obtained social support as a determinant of poor mental health.
 DETERMINANTS OF PERCEIVED SOCIAL SUPPORT 1143

Unfortunately, our modeling procedures cannot provide a test ceived obtained support at a given time can best be predicted
between these two important alternatives. However, the effects from the corresponding measure at the earlier time. The predic-
of reversing the causal influence on other estimates can be in- tive power of the earlier measure appears to be due to an associ-
structive with respect to the robustness of the model. When the ation produced by the causal influence of three other factors at
analyses were repeated on a model that reversed the direction both time periods. These factors are (a) the stability of provision
of influence from obtained social support to poor mental of support by the significant other, (b) the personality of the
health, the obtained results in terms of goodness of fit remained respondent in terms of a disposition to perceive himself and his
virtually unchanged, yielding a chi-square of 1378 with 704 de- interaction with the social environment (parents, others) in a
grees of freedom. There was only a small decrease in the betas negative light, and (c) poor mental health as indicated by anxi-
of the paths from obtained social support to poor mental health ety and depression. As hypothesized, when these three factors
(from -.18 and -.17 to -.06 and -.06 at Times 1 and 2, respec- were introduced into our final model (i.e., Figure 2), the beta
tively; cf. Figure 2). between obtained social support at Time 1 and Time 2 de-
We attempted a partial replication of the extended model in creased (from .69 in the initial model to .44 in our extended
Figure 2 by using the measures obtained at Time 2 as predictors model, Figure 2).
of their level at Time 3. The replication again indicated a high The extended structural model suggests that perception of
degree of goodness of fit to the model: The chi-square with 704 obtained social support reflects, rather substantially, the inter-
degrees of freedom was 1513, with the model having virtually personal transactions. At the same time, there is also consistent
identical beta coefficients with the same level of significance as evidence that obtained social support is influenced, though to
the earlier one in Figure 2. The ratio of the chi-squares to the a lesser degree, by the stable personality characteristic of a gen-
degrees of freedom was 2.15, and Hoelter's critical n index was eralized negative outlook.
216, indicating a good fit to the model. A generalized negative outlook appears to exert its strongest
In our final set of analyses we applied the extended model to influence on measures of affective states such as anxiety and
the data of the employed and the unemployed groups sepa- depression. There is also evidence for a small, but significant,
rately. The obtained chi square for the same model with 704 relation between poor mental health and the perception of ob-
degrees of freedom was 1083 and 1176 for the two groups con- tained social support. This relation is uncontaminated by the
secutively. The critical n goodness-of-fit values were 121 and effects of generalized negative outlook. Unfortunately, our
176, which indicated poorer fit of the data of these subgroups model does not allow us to determine the causal direction of
to the model than the fit of the data that was based on the entire this relation.
sample to the same model. Nevertheless, the corresponding beta Overall, the results of this study clearly demonstrate the need
coefficients produced by the analysis of the two subgroups were to consider both personality and social interaction when exam-
very similar to one another. They are not presented or discussed ining the nature of social support and its perception. Further-
separately because they displayed exactly the same pattern as more, the findings suggest that affective states also play a role in
for the combined data as in Figure 2. the perception of this interpersonal phenomenon.

Discussion Limilalions of the Study

Part of the interest in the concept of social support derives Despite the comprehensive analyses presented in this article,
from its potentially ameliorative effects on well-being. There is the results should be viewed with caution. The findings of our
evidence that it serves as a buffer between various stressors of analyses appear consistent and robust. Nevertheless, there are
life and well-being as well as has direct effects on well-being and four reasons why they do not provide complete and adequate
on reducing stressors (e.g., see reviews by Cobb, 1976; Cohen assessment of the role and the relative contribution of the con-
& Wills, 1985; Kessler & McLeod, 1985). As the literature has structs to perceived social support. First, the role of personality
grown, attention has turned to an important set of related is- was represented in this study by only a single construct, the
sues. These issues are concerned with the conceptualization of generalized negative outlook (although we believe that this con-
social support in terms of perceived availability versus actual struct is a key one). This construct was derived from available
interpersonal transactions and with the effects of personality on measures that were not specifically designed to tap the con-
the mobilization of social support and on its perception. struct. Hence it is quite plausible that we have omitted other
The study reported in this article focused primarily on two stable dispositions that influence the perception of interper-
of these issues: the extent to which perceived obtained support sonal interactions. For example, the disposition to respond to
reflects actual supportive transactions and the influence of per- events with intense affect (Larsen, Diener, & Emmons, 1986)
sonality on the perception of support. In the structural analyses may exert influence on the experience and perception of social
we attempted to highlight the mutual and relative influence of support. Various motivational and attributional dispositions,
three major variables on the perception of obtained social sup- particularly those that relate to other people, such as need for
port. The first variable, provided support, represents the contri- affiliation, power, or nurturance, are likely to affect the percep-
bution of the interpersonal transaction as reported by the pro- tion of supportive interpersonal interactions (e.g., Lefcourt,
vider of support. The other two variables, generalized negative Martin, Pick, & Saleh, 1985).
outlook and poor mental health, represent the influence of sta- Second, our measure of provided social support is likely to
ble and transient dispositions of the recipient of support. be influenced by stable personal dispositions of the provider in
The findings of the structural analyses with respect to ob- a manner similar to the way the measure of obtained support
tained social support are very consistent. They show that per- was shown to be influenced by the personality of the recipient.
1144 A. VINOKUR, Y. SCHUL, AND R. CAPLAN

Recall that both measures were based on perceptions of sup- set of questions regarding whether the perception of obtained
port: provided and obtained. This unmeasured influence may social undermining is influenced by the same parameters that
have biased the estimate of the link between provided and ob- influence the perception of obtained social support. For exam-
tained support. A more accurate assessment of the contribution ple, how much of the perception of obtained social support is
of provided support to obtained support could be achieved by influenced by provided social support compared with sent so-
modeling the determinants of the measure of provided support. cial undermining?
Third, the role of interpersonal transactions in this article was To take another example of the utility of examining social
assessed by the agreement in the provider's and the recipient's undermining, Rook's findings suggest that social undermining
report of social support. The literature already suggests that the is more likely than social support to influence mental health.
convergence in these measures provides only imperfect indica- Does this mean that the perception of social undermining is a
tion of the actual interpersonal transaction that transpired. In more salient and, accordingly, more influential element than
a study that used detailed behavioral checklists, Jacobson and social support? If so, would a generalized negative outlook be
Moore (1981) found only modest agreement between spouse more likely to influence the perception of negative social acts
reports of behaviors involving companionship, affection, sex, than of positive acts? Would negative mood states also be more
consideration, and communications. likely to elicit perceptions of social undermining than to lower
Fourth, whereas the recipients of social support in this inves- perceptions of social support? And, similarly, would sent social
tigation included only men, the social support providers in- undermining have a stronger effect on the perception of ob-
cluded primarily women (more than 87%). Gender differences tained social undermining than sent social support would have
in the pattern of social support that may have influenced our on the perception of obtained social support?
findings have been identified in a number of studies. For exam- We close by noting the practical value of pursuing these is-
ple, Butler, Giordano, and Neren (1985) found that female sues. There are a wide variety of applied settings in which the
graduate students reported both asking for and receiving more need to assess social support is potentially important. Interven-
support during periods of personal stress than did men. A more tions aimed at improving patient compliance by altering physi-
complete set of findings that is based on our extended model cian behavior (Eraker & Becker, 1984), at altering family pat-
should include women as well as men as the recipients of social terns of interaction (e.g., Minuchin, Rosman, & Baker, 1978),
support. and at changing the way in which managers and their employees
work together (Anderson, 1976), all rely on altering patterns of
social influence from the negative and undermining to support-
Implications/or Future Research
ive patterns. Experiments that involve such interventions and
The limitations of the study could be overcome if future stud- program evaluation must rely, at the outset, on a crucial manip-
ies provide more comprehensive and systematic assessments of ulation check of "what got delivered" and "what got received"
the relevant constructs. In addition, the modeling approach (Price, Detterer, Bader, & Monahan, 1982). As noted earlier,
demonstrated in this article could be applied to study the role some investigators' lines of research suggest that these self-re-
of two other factors in the interpersonal dynamics involved in ports will be influenced by the perceiver's mood. Other lines of
social support and its effects on well-being: the role of personal- research suggest that perceptual styles and biases will also be
ity in the mobilization of support and the role of social under- important. We have shown that both need to be assessed and
mining. controlled for so as to reveal a clearer picture of the link be-
As Heller and Swindler (1983) suggested, personality may tween "what got delivered" and "what got received."
play a crucial role in the mobilization of social support. Recent
work has already produced some evidence on the role of social References
skills and psychological hardiness in attracting social support
(Cohen, Sherrod, & Clark, 1986; Kobasa & Puccetti, 1983; Abbey, A., Abramis, D. J., & Caplan, R. D. (1985). Effects of different
B. R. Sarason et al., 1985; I. G. Sarason et al. 1986). To assess sources of social support and social conflict on emotional well-being.
Basic and Applied Social Psychology, 6, 111-129.
the contribution of personality to the mobilization of social
Anderson, J. (1976). Giving and receiving feedback. In P. R. Lawrence,
support, relevant personality variables can be integrated into
L. B. Barnes, & J. W. Lorsch (Eds.), Organizational behavior and
our model by linking them to the provided social support con-
administration (3rd ed., pp. 103-109). Homewood, IL: Dow Jones-
struct. Irwin.
Recent studies have suggested that social support and social Ban-era, M., Jr. (1986). Distinctions between social support concepts,
undermining need not necessarily fall on the same continuum measures, and models. American Journal of Community Psychology,
(e.g., Abbey, Abramis, & Caplan, 1985; Rook, 1984). These 14,413^*45.
studies suggest that some relationships may be high in both so- Berkman, L. E, & Syme, S. L. (1979). Social networks, host resistance,
cial support and undermining, others low in both, and others and mortality: A nine year follow-up study of Alameda County Resi-
high in one and low in the other. Social support scales, such as dents. American Journal of Epidemiology, 109, 186-204.
Blaney, P. H. (1986). Affect and memory: A review. Psychological Bulle-
the one we have used, focus only on supportive behaviors and,
tin. 99, 229-246.
consequently, cannot distinguish among these cases. The low
Blumenthal, M., Kahn, R. L., Andrews, F. M., & Head, K. B. (1972).
end of our social support scales merely represent inaction or
Justifying violence Ann Arbor, MI: Institute for Social Research.
ignoring someone rather than going out of one's way to under- Butler, X, Giordano, S., & Neren, S. (1985). Gender and sex-role attri-
mine them (Abbey et al., 1985). butes as predictors of utilization of natural support systems during
One useful extension of our model would be to introduce personal stress events. Sex Roles, 13,515-524.
social undermining as a separate variable. This would lead to a Caplan, R. D., Cobb, S., French, ;. R. P., Jr., Harrison, R. V., & Pin-
 DETERMINANTS OF PERCEIVED SOCIAL SUPPORT 1145

neau, S. R., Jr. (1975). Job demands and worker health: Main effects health in community samples. In S. Cohen & L. Syme (Eds.), Social
and occupational differences (USGPO No. HE7111-J57, USPO support and health (pp. 219-240). New "York: Academic Press.
Stock No. 1733-0083). Washington, DC: U.S. Government Printing Kobasa, S. C., & Puccetti, M. C. (1983). Personality and social resources
Office. in stress resistance. Journal of Personality and Social Psychology, 45,
Cobb, S. (1976). Social support as a moderator of life stress. Psycho- 839-850.
malic Medicine, 38, 300-314. LaRocco, J. M., House, J. S., & French, J. R. P., Jr. (1980). Social sup-
Cohen, S., Mermelstein, R., Kamarek, T, & Hoberman, H. N. (1985). port, occupational stress, and health. Journal of Health and Social
Measuring the funcional components of social support. In I. Sarason Behavior. 27,202-218.
& B. Sarason (Eds.), Social support: Theory, research and applica- Larsen, R. J., Diener, E., & Emmons, R. A. (1986). Affect intensity
tions (pp. 73-94). Dordrecht, the Netherlands: Martinus Nijhoff. and reaction to daily life events. Journal of Personality and Social
Psychology, SI, 803-814.
Cohen, S., Sherrod, D. R., & Clark, S. M. (1986). Social skills and the
Lefcourt, H. M., Martin, R. A., Fick, C. M., & Saleh, W. E. (1985).
stress-protective role of social support. Journal of Personality and So-
Locus of control for affiliation and behavior in social interactions.
cial Psychology, 50, 963-973.
Journal of Personality and Social Psychology. 48, 755-759.
Cohen, S., & Syme, L. S. (1985). Issues in the study and application of
Lefcourt, H. M., Martin, R. A., & Saleh, W. E. (1984). Locus of control
social support. In S. Cohen & L. Syme (Eds.), Social support and
and social support: Interactive moderators of stress. Journal of Per-
health (pp. 3-22). New York: Academic Press.
sonality and Social Psychology, 47, 378-389.
Cohen, S., & Wills, T. H. (1985). Stress, social support and the buffering
Lewinsohn, P. M., & Hoberman, H. M. (1982). Depression. In A. S.
hypothesis. Psychological Bulletin, 98, 310-357.
Bellack, M. Hersen, & A. E. Kazdin (Eds.), International handbook of
Coyne, J. C. (1976a). Depression and responses of others. Journal of behavior modification and therapy (pp. 397-430). New York: Plenum
Abnormal Psychology, 85, 186-193. Press.
Coyne, J. C. (1976b). Toward an interactional description of depression. Minuchin, S., Rosman, B. L., & Baker, L. (1978). Psychomaticfamilies.
Psychiatry, 39,28-40. Cambridge: Harvard University Press.
Derogatis, L. R., Lipman, R. S., Rickels, K., Uhlenhuth, E. H., & Covi, Price, R. H., Detterer, R. E, Bader, B. C., & Monahan, J. (1982). Preven-
L. (1974). The Hopkins Symptoms Checklist (HSCL): A measure of tion in mental health. Beverly Hills, CA: Sage.
primary symptom dimensions. In P. Pichot (Ed.), Psychological mea- Procidano, M. E., & Hellei; K. (1983). Measures of perceived social
surements in psychopharmacology. Modern problems inpharmacop- support from friends and from family: Three validation studies.
sychiatry(\<A. 7, pp. 79-110). Basel, Switzerland: Karger. American Journal of Community Psychology, I I , 1-24.
Eraker, S. A., & Becker, M. H. (1984). Improving compliance for the Rook, K. S. (1984). The negative side of social interaction: Impact on
patient with diabetes. Practical Diabelology, .J, 6-11. psychological well-being. Journal of Personality and Social Psychol-
Emmons, R. A., Diener, E., &Larsen, R. J. (1986). Choice avoidance of ogy, 46. 1097-1108.
everyday situations and affect congruence: Two models of reciprocal Sandier, I. N., & Lakey, B. (1982). Locus of control as a stress modera-
interactionism. Journal of Personality and Social Psychology, 51, tor: The role of control perceptions and social support. American
815-826. Journal of Community Psychology. 10,65-80.
Gold, M., & Mann, D. (1972). Delinquency as defense. American Jour- Sarason, B. R., Sarason, I. G., Hacker, T. A., & Basham, R. B. (1985).
Concomitants of social support: Social skills, physical attractiveness,
nal of Orthopsychiatry, 42, 463-479.
and gender. Journal of Personality and Social Psychology, 49, 469-
Gold, M., & Reimer, D. J. (1975). Changing patterns of delinquency
480.
behavior among Americans 13 through 16 years of age: 1967-1972.
Sarason, B. R., Shearin, E. N., Pierce, G. R., & Sarason, I. G. (1987).
Crime and Delinquency Literature, 7,483-517.
Interrelations of social support measures: Theoretical and practical
Heller, K., & Swindler, R. W. (1983). Social networks, perceived social
implications. Journal of Personality and Social Psychology, 52, 813-
support and coping with stress. In R. D. Felner, L. A. Jason, J. Morit- 832.
sugu, & S. S. Farber (Eds.), Preventive psychology: Theory, research, Sarason, I. G., Levine, H. M., Basham, R. B., & Sarason, B. R. (1983).
and practice in community intervention (pp. 87-103). New York: Per- Assessing social support: The social support questionnaire. Journal
gamon Press. of Personality and Social Psychology, 44, 127-139.
Hoelter, J. W. (1983). The analysis of covariance structures: Goodness- Sarason, I. G., & Sarason, B. R. (1986). Experimentally provided social
of-Fit indices. Sociological Methods and Research, 11,325-344. support. Journal of Personality and Social Psychology, 50, 1222-
House, J. S. (1981). Work stress and social support. Reading, MA: Ad- 1225.
dison-Wesley. Sarason, I. G., Sarason, B. R., & Shearin, E. N. (1986). Social support as
House, J. S., & Kahn, R. L. (1985). Measures and concepts of social an individual difference variable: Its stability, origins, and relational
support. In S. Cohen & L. Syme (Eds.), Social support and health aspects. Journal ofPersonality and Social Psychology, 50, 845-855.
(pp. 83-108). New York: Academic Press. Vinokur, A., Caplan, D. R., & Williams, C. C. (in press). Effects of
House, J. S., Robbins, C., & Metzner, H. L. (1982). The association of recent and past stresses on mental health: Coping with unemploy-
social relationships and activities with mortality: Prospective evi- ment among Vietnam veterans and nonveterans. Journal of Applied
dence from the Tecumseh Community Health Study. American Jour- Social Psychology.
nal of Epidemiology. 116, 123-140. Watson, D., & Clark, L. A. (1984). Negative affectivity: The disposition
to experience aversive emotional states. Psychological Bulletin, 96,
Jacobson, N. S., & Moore, D. (1981). Spouses as observers of the events
465-490.
in their relationship. Journal of Consulting and Clinical Psychology,
Wethington, E., & Kesslei; R. C. (1986). Perceived support, received
49, 269-277.
support, and adjustment to stressful life events. Journal of Health and
Joreskog, K. G., & Sorbom, D. (1984). L1SREL: Analysis of linear struc- Social Behavior, 27, 78-89.
tural relationships by maximum likelihood, instrumental variables,
and least squares methods. Chicago: National Educational Re- Received December 1, 1986
sources. Revision received June 22, 1987
Kessler, R. C., & McLeod, J. D. (1985). Social support and mental Accepted April 30, 1987 •