1“. ‘S‘fiffih‘! * v

.. y. .
#:3145933:
' - - ‘9- a

ii}?

173'

r,:f.:/-

,

2', ‘
xv

'E.

“2% ,,

Q‘.‘.

1;}:

* €21 ,glg
'l'" 74v;

1
..

.

g- ‘ .m
afi'i: it" s?” ,
,4 .l' -{',,7.

#5::

«53:3;

. ,
23'; $935.”
am, my: .»
4‘51“ '
4

‘V‘u'
3"?"
P3”

 

 

 

k H“.
’ ‘. "B‘fi;
’ ‘l
r!“ ml

4
hi ;
9% mi}:

 

 

SITY Liam

lllllllllllll||||l|llll|llllllllllllllllll

mas 3 1293 01399 462

 

llllllllll

LIIRARY
Michigan State
University

 

 

 

 

 

 

 

 

This is to certify that the

thesis entitled

Does Attachment Style Influence the Effects
of Social Support on the Stress—Illness
Relationship?

presented by

Matthew Paul Novak

has been accepted towards fulfillment
of the requirements for

Master ' 8 degree inmy

 

[Ar/W

Major professor

DatufléLQh—ZAhLQiQL

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

PLACE ll RETURN BOX to roman this cinchoui from your record.
TO AVOID FINES Mum on or before date duo.

DATE DUE DATE DUE DATE DUE

 

KAYJQ i 1907

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU is An Affirmative ActiorVEcpni Opportunity instituion
W

Does Attachment Style Influence the Effects of
Social Support on the Stress-Illness Relationship?

By

Matthew Paul Novak

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS
Department of Psychology

1995

ABSTRACT

Does Attachment Style Influence the Effects of
Social Support on the Stress-Illness Relationship?

By

Matthew Paul N ovak

This prospective study used path analysis to examine the effects of
attachment style on the stress-illness relationship as it acts through
social support. The subjects were 268 undergraduate students who
received class credit for their participation. It was found that two
factors of attachment, anxiousness about and avoidance of
relationships, did have a negative impact both on the size of and
satisfaction with a person's social support network. Social network
size was a significant predictor of stress at Time 1, thus supporting a
direct effects model. Physical symptoms at Time 1 were also found
to predict stress. However, the only significant predictor of
symptoms at Time 2 (4-6 weeks later) was symptoms at Time 1. The
moderating effects of social support on the stress-illness relationship

were not supported by the results of this study.

ii

DEDICATION
To Karen,
my wife and my best friend,
without whose love, support, and understanding

this work would not have been possible.

iii

ACKNOWLEDGMENTS

Thanks go to the following:

Robert Caldwell, Ph.D. for his guidance and patience throughout
the process of creating this work, and for making me do it on my
own.

Rick Deshon, Ph.D. for his expert assistance with statistics and
for his gentle instruction and patience.

Gary Stollak, Ph.D. for his support and guidance.

Richard and Kathleen Novak, my parents, for their love and
support.

Aaron Werbel, M.A., my friend, for listening to my seemingly
endless tribulations, and for his encouragement.

Derek Zmich, my friend, for assuring me that I was not alone in
how I felt.

Dave Gaultier, Mike Boardman, Claud Agnello, Bruce Roth, Rob
Curtner, Larry , Paul Emery, and Jim Shewell, the Monday night
group, for their friendly ears, kind hearts, and good counsel.

Jim Fuendeling, Jason Dahn, Josh Cohen, Frank DeMarco, Abby
Gleason, Jamie Jones, Steven Meyers, Regina O'Connell, and Usha
Tummala, my classmates and colleagues, for caring enough to listen
during my trials.

Carol Smiley, and Russ Smiley for their encouragement and

support, and for being there for me when I needed to bend an ear.

iv

John E. Hunter, Ph.D. for his statistical knowledge and
instruction.

Suzy Pavick, Cheryl Cook, Vicki Alexander, Leah Davis, Greta
McVay, and Ann Walker, the office staff, for their invaluable help in
putting together all of the paperwork needed for this study, and for
answering my many questions with kindness.

And to anyone else I may have forgotten to mention.

May God bless you all in whatever path you chose, as he has

blessed me with all of you.

TABLE OF CONTENTS

List of Tables
List of Figures
Introduction
Life events and physical health
Daily life events and physical illness
Social support and stress
Introduction
Direct effects of social support
Social support as a moderator
Possible extraneous variables
Why social support may not always be of help
Attachment and symptoms
Method
Subjects
Procedure
Materials
Results

Hypothesis 1

Hypothesis 2

Hypothesis 3

Hypothesis 4
Discussion

List of references

vi

vii

viii

OOOOLIIN

11
13
15
17
20
26
26
26
26
32
32
33
36
36
42
48

LIST OF TABLES

Table 1. Two factor solution with factor loadings. 30-31

Table 2. Means and standard deviations for all 3 2
variables.
Table 3. Correlation matrix for all variables. 3 3

vii

Figure 1.

Figure 2.

Figure 3.

Figure 4.

LIST OF FIGURES

A diagram of the path model which was
proposed for this study.

Model to test Hypothesis 2 with path
coefficients.

Model to test Hypothesis 4 with path
coefficients.

Modified path model with path coefficients.

viii

25

4O

1
Stress and Physical Illness

lutmducticn

The relationship between mind and body has been pondered at
least since the time of Hippocrates and his humoral theory. However,
scientific theories about mind-body interactions and the testing of
those theories are fairly recent developments.

In 1929 Walter Cannon produced some of the first scientific
evidence in support of the relationship between mind and body. He
labeled this the "Flight-or-Flight" response, and discovered that when
psychologically stressed, organisms show physiological responses.

In 1956, work in the area of mind-body connections was
furthered by Hans Selye in his pioneering book The Stress of Life .
Here Selye promoted his theory, the General Adaptation Syndrome,
to explain the mind-body connection we so familiarly speak of today.
The General Adaptation Syndrome consists of three stages. In the
first stage, called Alarm Reaction, the experience of a stressor, a
threatening situation, causes the mobilization of the body's resources.
Blood pressure rises, epinephrine, norepinephrine, and cortisol are
secreted by the endocrine system, heart rate increases, and blood
flow to the skeletal musculature is increased. However, this level of
arousal cannot be maintained for long without the organism being in
crisis. The Stage of Resistance follows. In this second stage, the body
attempts to adapt to the stressor. The level of arousal declines
somewhat from that of stage one, but continues at a higher than
normal rate. However, such a heightened state of arousal impairs the
organism from resisting new stressors or attacks. At this stage the

organism is susceptible to what Selye termed "diseases of

2
adaptation". Should the stressor continue beyond this point, the

organism will pass into the Stage of Exhaustion, marked by a

depletion of the body's energy reserves, extreme susceptibility to
disease, and physiological damage. Thus, the effects of prolonged
psychological stress could be detrimental to one's physical health.

In 1976, Cassel echoed Selye when he stated that "the
microbial diseases most common in our communities today arise
from the activities of microorganisms that are ubiquitous in the
environment, persist in the body without causing obvious harm
under ordinary circumstances, and exert pathological effects only
when the infected person is under conditions of physiological stress"
(p. 108). Since his address to the American Public Health
Association, much research has attempted to answer the question of
how psychological stress affects human health.

Selye (1956) defined stress as "the state manifested by a
specific syndrome which consists of all the nonspecifically induced
changes within a biologic system" (p. 54). While this biological
definition suited his work, for the purposes of this paper, stress will
be defined as the condition which results when a person perceives a
discrepancy between the demands of a situation and the resources at

his/her disposal (Sarafino, 1990).

'f v h
Much of the research done to investigate the relationship
between stress and physical health has focused on major life events -
prominent occurrences in one's life which have been identified by

consensus as stressful. This research has found that life events,

3
primarily negative life events, do have an adverse effect on physical

health (Dienstbier, 1989). Such stress does not appear to be illness-
specific (Gottlieb, 1985). Life events have been found to be related
to such varying diseases as upper respiratory tract infections,
allergies, increases incidence of hypertension, heart disease (Jemmott
& Locke, 1984), throat infections, tuberculosis, and influenza (Minter
& Kimball, 1978). The relationship also does not appear to be
dependent upon the type of stressor, as the studies cited examined
such disparate stressors as marital problems, death of a loved one,
change in housing, and having a child out of wedlock.

Many of the early studies investigating the effects of stress on
physical health did not report correlations between stress and health
or illness. But when correlations were reported they ranged from
r; = .12 (Rabkin & Struening, 1976) to r = .43 (Kobasa, Maddi, &
Puchetti, 1982), and were most often in the r = .205 range. Such
correlations were inconsistent and often of a low order.

Furthermore, the studies which examined this relationship often
used retrospective methods of data collection that rely on the
memory of the subjects. Such methods of data collection are often
subject to error and omission.

In addition, much of the early research in this area used self-
report methods to collect data, which may have created biases in the
results. Self-report data are problematic because people under
stress may become more sensitive to physical symptoms and adopt
the sick role without actually being ill (Minter & Kimball, 1978) and

because people experiencing negative life changes may selectively

4
remember more negative events, such as being sick. Recent research

has attempted to correct some of these problems.

For instance, to eliminate the subjective reporting of illness,
Jacobs and Charles (1980) examined life events in the families of 25
children with cancer and 25 other children brought to a clinic. The
families of the children with cancer reported experiencing
significantly more life events prior to diagnosis than did the controls.
This study improved on the efforts of its predecessors, but it is
possible that the life events may have been preceded by the onset of
the disease and therefore were the result of the children's illnesses
rather than the cause. And again, the retrospective nature of the
data collection may have skewed the results in favor of a significant
finding.

The most powerful studies have been prospective and have
used objective measures to assess health. For instance, Punch and
Marshall (1983) obtained stress data from 352 women with breast
cancer, and found that stress was a significant predictor of survival
20 years later. But even studies such as these have produced mixed
results. This may be due in part to studies which have looked at
single major life events.

Cohen and Wills (1985) suggested that "although a single
stressful event may not place great demands on the coping abilities
of most persons, it is when multiple problems accumulate, persisting
and straining the problem-solving capacity of the individual, that the
potential for serious disorder occurs" (cf. Wills & Langner, 1980).
Another explanation may be seen in the Illness Onset model of

disease. Proposed by Dorian and Garfinkel (1987), this model

5
suggests that acute stressful events do not have a primary effect on

one's health, but rather increase one's susceptibility to becoming ill
at a certain point in time. This is based on the belief that adaptation
is necessary to effectively cope with life change and that a lack of
adaptation may lead to alterations in one's immune functioning. In
their review of literature, Dorian and Garfinkel (1987) note a study
done by Dorian et a1. (1985) in which immune enhancement
followed the onset of a stressful situation (unemployment). It was
not until the situation became prolonged (chronic) that a decline in

immune function was noted.

'1 v i s

DeLongis, Coyne, Dakof, Folkman, and Lazarus (1982) suggested
that major life events may not cause stress in an individual's life
directly but may instead be stressful due to the way in which they
impact day-to-day living. They have proposed that life events are
"distal", that they have lost "specific functional significance" for the
individual, and therefore are described in nonpsychological terms
(DeLongis et al., 1982). Therefore, everyday disruptions in living
caused by major life events, called "hassles", would correlate more
highly with measures of health than major life events because
hassles are "proximal" events which often consist of "person-
environment interactions" that are particularly salient to an
individual and his/her well-being. The data supported their
hypotheses.

DeLongis et a1. (1982) found correlations between the

frequency of hassles and health status ranging from -.21 to -.33, with

6
the intensity of the hassles also correlating significantly with health

status (1;: -.29 to -.38). The only significant correlation reported
between life events and health status was between prestudy life
events and health status at the end of the study (1; = -.26). This
suggests that there is a delayed effect of stress on health. In order to
support their hypothesis that major life events affect health through
hassles, a multiple regression analysis was conducted. If hassles
were entered first, they accounted for 13% of the variance in overall
health status, with major life events not adding any significant
variance beyond this. And when life events were entered first, they
accounted for 7% of the variance, but hassles added an additional 9%
of the variance to health status. The analysis of the relationship
between hassles, life events, and somatic symptoms produced similar
results. Hassles were significantly correlated with symptoms (1; = .27
to .35), but life events were not. And a hierarchical multiple
regression demonstrated that hassles accounted for 13% of the
variance, with life events not adding significantly to this association.
When life events were entered first, they accounted for only 0.4% of
the variance, with hassles contributing an additional 14%. While
such results seem to lend overwhelming support to the hypothesis in
question, the study was retrospective. The authors therefore suggest
caution in interpreting their findings because it is possible that one's
health status may contribute to the hassles one reports.

Since this pioneering work, others have found evidence to
support its findings (Cox et al., 1984; Gannon, Banks, Shelton, &
Luchetta, 1989; Gannon, & Pardie,1989; Holahan, Holahan, & Belk,
1984 ; Weinberger, Hiner, & Tierney, 1987) with correlations

7
between hassles and symptoms ranging from .19 to .64. However,

some have found evidence contradicting this relationship between
minor everyday events and physical health.

Evans, Pitts, and Smith (1988) had 30 college students
complete the Assessment of Daily Experiences and a symptom
checklist each day for nine weeks. They found that there was a
significant decline in desirable events in the four days preceding an
illness. However, the subjects of their study reported no significant
change in undesirable events in the same time period. The authors
suggest that it is the "absence of uplifts rather than the presence of
hassles [which] seems to be the danger signal" (p. 539) prior to an
illness episode. Similar results have been obtained by others (Stone,
Bruce, & Neale, 1987; Zarski, 1984), but the preponderance of
evidence appears to support the notion that chronic stressful events

such as daily hassles are correlated with physical symptomology.

8
Social support and stress

1111mm

The modest findings of stress' effects on physical health have
led psychologists to search for other factors that may be involved in
the process. In 1976, John Cassel stated that "it is evident that any
disease process, and in fact any process within the living organism,
might be influenced by the reaction of the individual to his social
environment or to other people" (p.109). The area of social support
seemed to promise to explain some of the inconsistencies in stress-
illness research. This section will present research exploring social
support's effects on the stress-illness relationship.

For the purposes of this discussion, social support will be
defined as "1. Information leading the subject to believe that he [sic]
is cared for and loved. 2. Information leading the subject to believe
that he is esteemed and valued. 3. Information leading the subject
to believe that he belongs to a network of communication and mutual
obligation." (Cobb, 1976, p.300)

In their pioneering work Berkman and Syme (1979) looked for
a connection between social support and illness. They examined
4725 people (2229 men and 2496 women) who had previously taken
part in the 1965 Human Population Laboratory survey in Alameda
County, California. The size (a measure of structure) of each subject's
social support network at the time of original data collection was
negatively related to mortality during the following nine years.
After adjusting for age and prior health status, men with low levels
of social support were 2.3 times more likely to die than those with

high social support. For women the increased risk was 2.8. The

9
results remained significant at the p<.001 level even after

simultaneously controlling for smoking, obesity, alcohol consumption,
physical activity, and age. Berkman and Syme also suggested that
since ill people with poorer social networks did not die in the first
few years of this follow-up study, then physical illness could not
account for all of the connection between social isolation and
mortality.

While the Alameda County Study was ambitious in its shear
size, it had many flaws which have since been addressed by other
studies. For instance, while it was prospective in nature, it assumed
that the data collected at the beginning of the study remained
consistent for the subjects during the nine subsequent years. Also,
there was poor control for factors such as physical health at the
outset of the study.

House, Robbins, and Metzer (1982) corrected some of these
flaws in their study of 1322 men and 1432 women. They collected
social integration and medical examination data during a ten year
period (1959-1969). Mortality from 1978-1979 was the dependent
variable. A wide variety of factors including prior health status,
lifestyle, and SES were controlled. Cumulative indices of social
relationships and activities were significant predictors of mortality
for men. For women, however the significant relationship between
cumulative indices of social integration became nonsignificant when
other factors were entered as covariates. They found no evidence to
support the proposition that satisfaction with relationships and
activities has a significant impact on mortality. Prior health status

did not appear to be related to social support. That is, those who

1 0
were ill did not necessarily have lower scores on social support

indices.

In addition to mortality studies, decreases in social support
have also been associated with increases in physical symptoms (see
Broadhead et al., 1983 for a detailed review). High levels of social
support have been related to lower risks for heart disease (Cohen,
1988; Siegrist, Siegrist, & Weber, 1986), decreases in low birth
weight, and better recovery from tuberculosis, asthma, heart disease,
and surgery (Cobb, 1976). The effects of social support do not
appear to be stressor specific, as social support has been noted to be
beneficial in alleviating the effects of a wide variety of stressors such
as unemployment, bereavement, and retirement (Cobb, 1976). But
there are enough negative findings to "make it clear that social
support is not a panacea" (Cobb, 1976, p. 310).

Poor social support may lead to anxiety and depression, which
may affect health through increasing susceptibility to disease, or
indirectly through changes in behavioral patterns that increase one's
risk for disease (Cohen & Wills, 1985). It has been suggested that
there are different types of social support including esteem support,
informational support, social companionship, and instrumental
support (Cohen & Wills, 1985). However, the four types are often so
highly correlated that the distinction between them becomes
unimportant (Caldwell & Reinhart, 1988; Stokes & Wilson, 1984).
But, the amount, type, source, and structure of social support are all
important dimensions in the defining the effectiveness of one's social

support network (Thoits, 1982).

l 1
Research in the area of social support has led to the proposal of

two models of social support: (a) the buffering model and (b) the
main effect model (Cohen & Wills, 1985). The main effects model
holds that social support is beneficial regardless of whether one is
stressed, while the buffering model suggests that support only works
for people while they are under stress. In general "evidence for a
buffering model is found when the social support measure assesses
interpersonal resources that are responsive to the needs elicited by
stressful events. Evidence for a main effect model is found when the
support measure assesses a person's degree of integration in a large
community social network" (pp.347-348). Put more simply, the
buffering model is usually supported when one examines one's
satisfaction with his/her support system, and main effects generally

found when measuring the structure of one's support network.

f i l r

One way in which social support eases the effects of stress on
one's body is by protecting the individual from experiencing stress in
the first place. This model is tested by examining the effects of social
support on symptoms without taking into account the level of stress
experienced by the subject. It has been suggested that it is the
structure of one's support system which best demonstrates the main
effects model (Cohen & Wills, 1985). That is, the better the structure
(a measure which may include dimensions such as size, density, and
integration) of one's network, the fewer symptoms one is likely to

experience.

1 2
The main effects model of social support was supported in a

study by Gore (1978). She compared 100 unemployed men to 74
employed controls over a period of two years to determine the
effects of social support in alleviating the effects of the stress on
physical health. Men with less support who became unemployed
reported more symptoms of illness than men who were more
supported or who remained employed. More important in light of
the main effects model was that "the well supported report[ed] lower
mean levels of symptoms than [did] the quickly reemployed" (p.162).
However, the author cautions that since the study did not include
adequate baseline levels (the study was started during anticipation
of unemployment, a stressor in itself), the results are equivocal.

Monroe (1983) corrected this error in design. He collected life
events and physical symptom data from seventy employees of a
corporation. While retrospective analysis showed that social support
was not significantly related to physical symptoms, prospective
analysis showed that support was a significant predictor of
subsequent symptoms regardless of stress level.

However "if social support [is] to be considered [an]
independent predictor of disorder, [it] should demonstrate an
association with subsequent symptoms once initial symptoms have
been taken into account" (Monroe, 1983, p.83). Monroe (1983) found
evidence to support this. Social support continued to contribute
significantly to the variance in physical symptoms even after prior
physical status was taken into account. This adds further support to

the main effects model.

1 3
According to Cohen and Wills (1985) the main effect model

may work because "large social networks provide persons with
regular positive experiences and a set of stable, socially rewarded
roles in the community" (p.311) thus providing a "positive affect, a
sense of predictability and stability in one's life situation, and a
recognition of self-worth" (p.311), or by decreasing stressful
experiences. Such factors may affect the immune or neuroendocrine
systems, or may influence one's health-related behaviors.

While there is some evidence to support a main effects model,
there is also evidence to suggest that the quality of one's social
support network is a better predictor of health than the structure of
that network (Broadhead et al., 1983). "When structural measures of
social network size are used to indicate the benefits of social
relationships, two questionable assumptions are made, namely, that
any benefits are directly proportional to the size and range of the
network and that having a relationship is equivalent to getting

support" (Schaefer, Coyne & Lazarus, 1981, p.383).

W

The buffering effect of social support occurs after a stressful
event has been experienced and is evident when there is an
interaction between levels of stress and levels of social support. That
is, people with different levels of support may report the same
number of symptoms when stress levels are low, but when stress
levels are high, those people with more social support report fewer
symptoms than those with low levels of support. The buffering

effect is most frequently noted when an experimenter is measuring

14
the functioning of the social support (e.g. satisfaction with social

support) rather than size or structure. Often such measures involve
satisfaction with a small subset of one's social network such as a
confidant, spouse, or family (Cohen & Wills, 1985).

A study conducted in Finland found that high school students
who had reported both a high level of life events and poor
relationships with their families and friends reported more
symptoms than students with good relationships or low levels of
stress (Aro, Hanninen, & Olavi, 1989). However, this relationship was
found to be significant only for boys.

Similar effects were found in a study of residents of Three Mile
Island (TMI) after the nuclear accident that happened there. TMI
residents with moderate or high levels of social support had
symptom levels comparable to control subjects with similar support
(Fleming, Baum, Gisriel, & Gatchel, 1982). Buffering effects have also
been implicated in research investigating a variety of health
conditions including herpes simplex virus (VanderPlate, Aral, &
Magder, 1988), and coronary heart disease (reviewed by Cohen,
1988).

However, these studies were done retrospectively and
therefore the results are questionable. But such effects have also
been noted in prospective research. For example, Monroe (1983)
found that subjects with initially high levels of symptoms and social
support tended to experience fewer symptoms during follow-up.
High levels of life events and low levels of social support together

were most predictive of high levels of physical symptoms, while high

15
support among those who experienced large numbers of life events

demonstrated the buffering effect.

Similar results were also found by Cohen and Hoberman
(1983). Using prospective measures of support they found support
for the buffering model. Interestingly, they found that when they
used retrospective measures, they did not find similar support. The
authors suggested that retrospective measures may not be good
measures of social support because they may tap a person's prior
need for support as well as its availability.

There is some evidence to suggest that social support may
reduce psychological stress, but does not do particularly well in
reducing physiological arousal to stress (Fleming, Baum, Gisriel, &
Gatchel, 1982). But support for the buffering model has also been
found in research which used objective measures of symptoms (e.g.,
Fontana, Kems, Rosenberg, & Colonese, 1989; Norbeck & Tilden, 1983;
Sarason, Sarason, Potter, & Antoni, 1985).

The buffering effects of social support may occur by (a)
preventing a "stress appraisal response" (i.e., may allow a person to
believe that s/he has sufficient resources to handle the stressor due
to social support's added resources) or (b) by reducing or eliminating
a stress reaction after its onset (through providing a solution to the
stressor, reducing the importance of the event, or by encouraging

healthful behaviors) (Cohen & Wills, 1985).

e V ' 1
It should be noted that there may be some other variables

which affect the stress-illness relationship. A study conducted by

16
McFarlane, Norman, Streiner, and Roy (1983) found that the number

of neutral or undesirable events experienced were associated with
physical health when measured concurrently. But they proposed
that "while stressful events may be related to consistent patterns of
health, they may not predict change in health. Alternatively,
individuals suffering from poor health may, as a result, experience
an increase in stressful events" (p. 165). After controlling for prior
health, they found that stressful events which occurred in the
previous six months were not related to current health, although
there was still a significant relationship between current life events
and current illness. Prior illness was significantly correlated with
current life events. They concluded that "the more immediate effect
of stressful events on health has been considerably over-estimated"
(p. 169), and offer three possible explanations for their finding that
stress is related to subsequent health problems: (a) Stressful events
may lead to later health problems, (b) symptoms may lead to higher
levels of stress rather than the other way around (but they note that
this does not explain the baseline level of stressful events reported
by the subjects of this experiment), and (0) some third factor, such as
personality, may cause both stressful events and poorer health.

Other possible extraneous variables are found under the
heading of lifestyle. People have good and bad habits which affect
health and may be unrelated either to attachment style or social
support. The most well known of these is diet.

A balanced diet consisting of meats, cereals, vegetables, fruits,
and dairy products is important to maintaining good health (Katch &

McArdle, 1993). Poor diet is associated with many health problems

17
including cardiovascular disease (Havel, 1989), hypertension (Kaplan,

1989), diabetes (Wolever & Jenkins, 1989), and cancer (Sugimura,
Wahabayashi, Nagao, & Ohgaki, 1989). Exercise also seems to
promote good health, apparently through improving immune system
functioning (Keast & Morton, 1992; Verde, 1992).

In addition to behaviors which may improve a person's health,
there are behaviors which may lead to an increased incidence of
illness. Alcohol abuse is correlated with various health problems
including nervous system disorders (Bigby, 1987), liver disease
(Moulton & Cyr, 1987), gastrointestinal problems (Cyr & Moulton,
1987), and cardiovascular disease (Ende, 1987). Likewise, people
who abuse drugs tend to have a higher incidence of illness than those

who do not use drugs (Andersen, 1981).

W ' N w 1

While there has been much research supporting social
support's beneficial effects on health, many criticisms have been
raised surrounding this research. It has been proposed that people
with strong networks may be more prevalent in studies which
examine such disorders such as coronary artery. disease because
their social networks will encourage them to seek treatment (Cohen
& Matthews, 1987), thus suggesting that people who most need social
support receive it.

One study found that while those with higher levels of social
support had less trouble with their ability to carry out ordinary daily
activities and maintain relationships than the others, when physical

status and health were entered first, social support did not.

1 8
contribute significantly to the variance of this level of functioning

(Lee, Graydon, & Ross, 1991). Similar results have been found in
other studies (Cox et al., 1984; Gannon & Pardie, 1989; Schaefer,
Coyne, & Lazarus, 1981).

One possible explanation for negative findings in the area of
social support was reported by Cohen and Hoberman (1983).
Although demonstrating social support's buffering effect, they found
a decrease in symptoms for those with low stress and low levels of
social support. This is not typical of evidence supporting the
buffering model. They hypothesized that "the increased
responsibilities of the interpersonal relationships may
themselves contribute a small increment to one's stress level and
consequently in symptomatology" (p.116).

Schaefer, Coyne, and Lazarus (1981) agree with Cohen and
Hoberman. They state that the assumption that network size and
amount of social support are positively associated "ignores the
demands, constraints, and conflicts also associated with social
relationships", which "comprise a significant share of the stresses
people experience in their daily liveS" (PP. 383-4). They further
propose that these "demands and constraints of network
membership may dilute or vitiate the beneficial effects, and so
weaken any relationship with health outcomes" (p.385). Such social
interaction would also be harmful if such interaction were improper
and would thus promote illness rather than health (Orth-Gomér &
Undén, 1987).

One study (McFarlane, Norman, Streiner, & Roy, 1983) found an

inverse relationship between stressful events and social supports.

19
The authors suggested that the size of one's social network may be

related to the perception of it being less helpful. Another possibility
is that increases in stressful events lead to the perception that one's
social network is less helpful.

But there is some evidence to refute such ideas. Fontana,
Kerns, Rosenberg, and Colonese (1989) tested the hypothesis that
illness, distress, and stress mobilize one's social support system to
action against the hypothesis that illness decreases support. The
data revealed no significant interactions. Therefore, there was no
support for buffering or moderating effects. They did, however, find
that intimacy ameliorated psychological distress and physical
symptoms, thus supporting the main effects model.

It may be that the positive health effects of social support are
countered by other factors. One's social network may decrease in
size because the demands of illness restrict the amount of time and
energy some people have to spend with the members of their
networks (Bloom, 1990). Or disease may decrease social support
because the sufferer overuses the system while coping with his/her
illness (Cohen & Matthews, 1987).

Finally, some people may be better able to use their social
support systems when faced with a stressful situation. Research in

the area of attachment theory seems to point in this direction.

20
Attachment and Symptoms

In 1958 John Bowlby proposed a theory explaining
interpersonal behavior. This theory stated that very young children
develop behaviors which maintain proximity to their primary
caregiver. Bowlby called these behaviors attachment behaviors.
Over time, the child develops internal representations of the
relationship with his/her caregiver. And eventually they become a
pattern of expectations for all interpersonal interactions which color
the individual's interactions with others. The tone of the patterns of
expectations is based on the success or failure of these behaviors and
the responsiveness of the caregiver.

Mary Ainsworth and her colleagues (1978) furthered Bowlby's
work and found three broad patterns of infant attachment to
caregivers. The first pattern has been termed secure attachment.
Those in this category appear to experience his/her primary
caregiver as accessible and responsive, and are characterized by
being easily consoled when distressed, compliant and cooperative,
outgoing, enthusiastic, positive, and persistent. Children displaying
the second pattern, the anxiously or ambivalently attached, tend to
have a less consistently positive experience of their caregivers, and
are characterized as anxious, clingy, easily frustrated, dependent,
passive-aggressive, and are not as easily soothed as secure children.
Finally, the avoidantly attached often experience his/her primary
caregiver as rejecting, and tend to be angry, aloof, and stand-offish.

Research over the past decade has supported these models.
Infants labeled secure children have shown a greater ability to

respond flexibly, persistently, and resourcefully in a preschool

2 1
setting (Arend, Gove, & Sroufe, 1979; Sroufe, 1983) and it appears

that avoidant children cut off anger or distress-related emotions
(Kobak, 1986; Lutkenhaus, Grossmann, & Grossmann, 1985). Thus,
experiences which lead to internal representations of others who are
unavailable lead to dysfunctions of attachment behavior (Sheldon &
West, 1990).

It is believed that these patterns "will be persistent and
resistant to change" (Mrazek, Casey, & Anderson, 1987, p. 516).
There is research supporting this assertion. A longitudinal study
done in Australia demonstrated a correlation between early
attachment experiences with parents and later marital quality
(Kotler & Omodei, 1988).

Simpson, Rholes, and Nelligan (1992) stated that:

The utility of Bowlby's attachment theory for adult

relationships, however, does not require that attachment styles

observed in adulthood date back to infancy. As long as the
patterns of attachment identified in children are
phenotypically similar to those that characterize adults, and as
long as the consequences of these styles for behavior and
emotions are similar across different developmental levels,
attachment theory remains a viable model for understanding

adult relationships (p. 443).

Research suggests that securely attached adults tend to report
little distress and high levels of social support (Kobak & Sceery,
1988). They feel valued and worthy of others' concern, support, and
affection, and develop mental models of significant others who are

accessible, reliable, trustworthy, and well-intentioned (Simpson,

22
Rholes, & Nelligan, 1992). Securely attached adults tend to express

trusting attitudes towards others (Feeney & Noller, 1990) and to be
involved in romantic relationships marked by higher levels of
satisfaction (Simpson, 1990).

Avoidant individuals picture themselves as aloof, and
emotionally distant, feel uncomfortable being around others, and find
it difficult to completely trust and depend on others (Simpson,
Rholes, & Nelligan, 1992). They report more distant relationships
(i.e., more loneliness) and lower levels of social support than the
other two groups (Kobak & Sceery, 1988).

Ambivalently attached persons report higher levels of distress
than the securely attached (Kobak & Sceery, 1988), express
dependence (Feeney & Noller, 1990), and feel that they are
misunderstood and underappreciated (Simpson, Rholes, & Nelligan,
1992). While they view their social network as more supportive
than the avoidantly attached (Kobak & Sceery, 1988), they believe
that others are reluctant to get as close as they wish (Simpson,
Rholes, & Nelligan, 1992).

Bowlby (1969) has argued that an individual's attachment
system should be most strongly activated under conditions of
distress. There have been several studies supporting this assertion.
For example, Simpson, Rholes, and Nelligan (1992) found that
securely attached women were more likely than avoidant women to
seek out support from their romantic partners when made to feel
anxious. However, no significant effects were found for the anxious

attachment style. Therefore,

23
secure attachment would be organized by rules that allow

acknowledgment of distress and turning to others for support,

avoidant attachment by rules that restrict acknowledgment of

distress and the associated attachment attempts to seek
comfort and support, and ambivalent attachment by rules that

direct attention toward distress and attachment figures in a

hypervigilant manner that inhibits the development of

autonomy and self-confidence (Kobak & Sceery, 1988, p. 142).
Simply stated, because securely attached people are relatively
comfortable with social relationships, they tend to have large social
networks and tend to be more satisfied with their networks.
Alternately, those people who are ambivalently or avoidantly
attached are less comfortable with social relationships, they tend to
have small networks and report low levels of satisfaction.

There is also evidence to indicate that attachment style is
correlated with physical illness. Mrazek, Casey, and Anderson (1987)
found that asthmatic children were more likely to be insecurely
attached than children without asthma. It is possible, therefore, that
attachment style moderates the effects of stress on physical health
through its effects on how one develops and uses his/her social
support system.

This study was designed to examine the effects of social
support on the relationship between stress and physical illness and
how attachment style might influence these effects. The following

hypotheses were tested:

24
Hypothesis 1: The amount of stress one experiences causes the

amount of physical illness experienced, after accounting for lifestyle
and prior illness.

Hypothesis 2: Given the same amount of stress, those persons
with relatively higher levels of social support experience less
physical illness than those with relatively lower levels of social
support. The effects of social support will be evident in two ways:
One, people with a higher number of people in their social support
networks will show less stress (i.e. direct effects model of social
support) and, two, people who are more satisfied with their social
support networks will have a smaller number of physical symptoms
even though their stress levels may be similar to those who are
unsatisfied with their networks (i.e. moderating model of social
support).

Hypothesis 3: A person's level of secure attachment is
positively correlated with the size of his/her social network as well
as with the level of satisfaction with the support received from said
network. Alternately, one's level of anxious and avoidant attachment
(as defined by Hazan and Shaver, 1987) is negatively correlated with
a person's social network size and the satisfaction received from that
network.

Hypothesis 4: This hypothesis proposed the path model
depicted in Figure 1.

Based on the literature reviewed in this paper, there is a
correlation between stress and health, social support can have a
direct effect on illness, or it can moderate the stress-illness

relationship, and attachment style appears to be related to health

25
and social support. Therefore, the above model purports that, after

accounting for prior illness and lifestyle factors, attachment will
moderate the relationship between stress and physical illness by

acting through social support.

 

 

social
suppon
satisfaction

 

l SGCUI’B

 

 

 

I ambivalent

 

 

social -
support

size ~

physical

1 ' illness

 

I avoidant

 

 

 

 

   

 

 

 

 

 

 

 

pfior
inness

 

Figure 1. A diagram of the path model which was proposed for this
study. (The plus or minus sign next to each proposed path indicates
the expected direction of that relationship.)

2 6
Method

Subjects

The subjects for this experiment 304 undergraduates enrolled
in introductory psychology classes at Michigan State University. In
exchange for their voluntary participation, they were given class
credit and were debriefed as to the nature of the experiment after
they completed the questionnaires. Twenty-one did not return for
the second testing session. In addition, fifteen surveys were
incomplete, and, due to the nature of the analysis, these data were

not included in the analysis. The final number of valid protocols was

268; 105 males and 163 females.

Emeritus

The subjects were tested at two times. The second testing
occurred four to six weeks after the first time. At Time One they
were asked to complete a lifestyle assessment questionnaire, the
Hassles scale, the Social Support Questionnaire, the Relationship Style
Questionnaire, and the Cohen-Hoberman Inventory of Physical
Symptoms. At Time Two they were asked to complete again the
lifestyle assessment questionnaire, and the Cohen-Hoberman

Inventory of Physical Symptoms.

Materials

The lifestyle assessment questionnaire is a 34-item self-report
measure derived from the Olin Health Center Lifestyle Assessment
questionnaire. It was designed to assess various lifestyle factors

such as eating, exercise, smoking, and self-care habits, as well as

27
alcohol and drug use patterns and sexual health. The subject

responds to questions on a five-point Likert scale. This measure was
used to control for lifestyle factors which may have influence the
stress-illness relationship. A lifestyle rating was obtained by adding
the responses for each question with a higher score indicating a more
healthful lifestyle. For this sample alpha reliability was .82 and test-
retest reliability was .87.

The Hassles Scale (Kanner, Coyne, Schaefer, & Lazarus, 1981) is
a 117-item self-report measure which assesses the "frequency and
severity of a person's transactions with the environment that are
considered by the person to be stressful events" (Lazarus & Folkman,
1989, p.1). In this experiment it measured perceived stress by
summing up the severity for all items endorsed. The authors of the
Hassles Scale reported test-retest reliability for severity was .48,
which is an acceptable level for a state measure. An alpha reliability
analysis was done using the sample chosen for this study and was
found to be .96. The scale also appears to have face and content
validity (Lazarus & Folkman, 1989).

The Social Support Questionnaire (Sarason, Levine, Basham, &
Sarason, 1983) is a 27-item self-report measure which asks the
respondent for the number of people to which s/he can turn in a
variety of situations. Each situation also includes a 6-point Likert
scale which asks about the subjects satisfaction with his/her support
system. Overall number and satisfaction scores were computed by
summing the number and satisfaction portions of each item,

respectively. Test-retest reliability with an undergraduate

28
population was .90 for number and .83 for satisfaction (Sarason,

Levine, Basham, & Sarason, 1983).

The Relationship Style Questionnaire (Bartholomew, 1990) is a
self-report measure which is based on a three previously designed
measures: The three-item measure by Hazan and Shaver (1987)
divided into its constituent statements, the four-style Relationship
Questionnaire (Bartholomew & Horowitz, 1991), again, divided into
its constituent statements, and Collins and Read's (1990) Adult
Attachment Scale. This questionnaire asks the subject to rate on a 5-
point Likert scale 30 statements as to how "like me" each is.
Originally it was intended that each subject would receive a score for
each of Hazan and Shaver's attachment categories by adding their
scores on statements corresponding to each category. However, it
was demonstrated that this method was undesirable. Although the
alpha reliability for the Ambivalent and Avoidant categories was
adequate (.71 and .72, respectively), reliability for the Secure scale
was only .60. Since others (Simpson, 1990; Simpson, Rholes, &
Nelligan, 1992) have reported that a factor analysis of similar scales
resulted in two factors, a forced two-factor exploratory factor
analysis was done. The factor analysis resulted in two factors which
closely resembled what have been labeled by others as Anxiousness
and Avoidance (Simpson, 1990; Simpson, Rholes, & Nelligan, 1992).
After modifying the factors to more closely fit these labels, alpha
reliability was .85 for Anxiousness and .81 for Avoidance. The items
contained in each factor are presented in Table l. Hypothesis 3 was,
therefore, restated to take into account this change in factors: A

person's level of anxiousness about and avoidance of close

29
relationships is correlated with the size of his/her social network as

well as with the level of satisfaction with the support received from
said network. That is, the higher one's levels of anxiousness and
avoidance, the smaller the person's network and the less satisfied
they will be with said network (see Figure 2).

Finally, the Cohen-Hoberman Inventory of Physical Symptoms
(Cohen & Hoberman, 1983) is a 33-item self-report measure which
asks the subject how much each item bothered or distressed the
subject. Items are rated on a 5-point Likert scale ranging from "not
at all" to "extremely". Reliability for this measure was .90 which is
similar to what Cohen and Hoberman found (alpha = .88). Test-retest
reliability was .66 for the sample used in this study. A physical
symptom score was derived by adding the responses from all items.

Therefore, a higher score indicated more symptoms reported.

Table 1

flfwu tagtur solutiuu with tagtut luauings. (An * indicates that this

item was moved from the other factor.

30

was deleted from the factor to increase reliability.

A § indicates that the item

An "r" indicates

 

 

an item which is reverse scored.)
RSO Item Fagtur 1 Eactut 2
Factor 1: Anxiousness
184. I often worry that romantic partners won't .77 .13
want to stay with me.
186. I worry about being abandoned. .72 .04
174. I often worry that romantic partners don't .66 .25
really love me.
172. I worry about being alone. .66 -.01
179. I worry that others don't value me as much .65 .16
as I value them.
181. My desire to merge completely sometimes .60 -.17
scares people away.
188. I find that others are reluctant to get as .57 -.05
close as I would like.
191. I worry about having others not accept me. .54 .04
180. People are never there when you need .52 .35
them.
168. I worry that I will be hurt if I allow myself .52 .38
to become to close to others.
§l67. I want to merge completely with another .49 -.40
person.
*170. I am not sure that I can always depend on .41 .43
other to be there when I need them.
*r190. I know that others will be there when I 18 .56

need them.

31
Table 2 (continued)

flfwo tactut sulutiun with factor luadings. (An * indicates that this

item was moved from the other factor. A § indicates that the item
was deleted from the factor to increase reliability. An "r" indicates
an item which is reverse scored.)

 

 

me.

RSO Item flictor 1 Fm;
Factor 2: Avoidance
r193. I find it relatively easy to get close to . .01 .65
others
r173. I am comfortable depending on other .15 .62
people.
164. I find it difficult to depend on other people. .30 .61
183. I am nervous when anyone gets too close to .38 .58
me.
189. I prefer not to depend on others. .11 .56
r166. I find it easy to get emotionally close to . -.15 .54
others
176. I worry about others getting too close to .29 .53
me.
175. I find it difficult to trust other completely. .45 .52
r177. I want emotionally close relationships. .40 -.52
187. I am somewhat uncomfortable being close .33 .47
to others.
192. Romantic partners often want me to be .09 .47
closer than I feel comfortable being.
r171. I want to be completely emotionally .42 .46
intimate with others.
165. It is very important to me to feel -.05 .42
independenL
169. I am comfortable without close emotional -.21 .33
relationships.
r178. I am comfortable having other people .01 .33
depend on me.
182. It is very important to me to feel self- -.01 .26
sufficient.
185. I prefer not to have other people depend on .26 .26

3 2
Results

The means and standard deviations for each measure are listed

in Table 2 and the correlations between variables are listed in Table

3.

Hyputhesis 1

The mean scores for Symptoms at Time 1 (8X1) and Symptoms
at Time 2 (SX2) were 56.28 (s.d. =17.15) and 59.14 (s.d. = 22.33),
respectively. The mean for Stress (STR) was 188.41 with a standard
deviation of 40.53. The mean score for Lifestyle at Time 1 (LFS) was
175.32 with a standard deviation of 14.98.

 

Table 2
e n r viin fr llva'abls.
Variable Mean S. Dev.
Anxiousness 31.97 9.65
Avoidance 46.80 10.32
SS Satisfaction 140.01 18.56
SS Number 104.59 39.57
Lifestyle Tl 175.24 14.95
Stress 188.41 40.53
Symptoms T1 56.28 17.15
Symptoms T2 59.14 22.33

Multiple regression on SPSS was used to test Hypothesis 1.
Symptoms at Time 2 was the dependent variable with Stress,
Lifestyle at Time 1, and Symptoms at Time I entered as predictor
variables. SXl was significantly positively related to SX2 (B: .62, p<
.05). However, while both LFS and STR were significantly correlated
with SX2 when either was taken in isolation (F -.25, p<.05 and 1;:

.46, p<.05; respectively), neither was significantly related to SX2 after

33
8X1 was accounted for (B: -.09 and B: .04, respectively). Therefore,

the best predictor of Symptoms at Time 2 was Symptoms at Time 1.

Thus, hypothesis 1 was not supported by these data.

Table 3
Cunelatiuu mattia fut all vmables (N=268). ( L > .12, p < .05; L >.16,

p < .01) (ANX=Anxiousness, AVO=Avoidance, SSN=Social Support
Number, SSS=Social Support Satisfaction, LFS=Lifestyle at Time 1,
SX1=Symptoms at Time 1, STR=Stress, SX2=Symptoms at Time 2)

ANX AVO SSN SSS LFS SXl STR SX2
ANX 1.00
AVG .37 1.00
SSN -.24 -.28 1.00
SSS -.35 -.28 .39 1.00
LFS -.31 -.12 -.Ol .08 1.00
SXl .38 .17 .01 -.07 -.24 1.00
STR .45 .24 -.11 -.09 -.3O .63 1.00
SX2 .29 .09 .Ol -.02 -.25 .66 .461.00

3199mm

To test whether Social Support Number (SSN) had a direct and
Social Support Satisfaction (SSS) had a moderating influence on the
STR-SX2 relationship, SSN and SSS, along with SXl and LFS, were
entered into a path analysis using PATH (Hunter & Hamilton, 1992)
(see Figure 2). The mean Social Support Number was 104.59 with a
standard deviation of 39.57, and the mean Satisfaction with Social
Support was 140.01 with a standard deviation of 18.56.

The direct effect of SSN on STR was not supported by the path
analysis (B: -.11, p> .05), although the path was close to being
significant. The moderator effect of SSS on the STR-SX2

34

 

 

 

 
  
  
     
   
   

 

ass |
.03(07) ,

 

 

 

 

-.11(.06) '°3('°7)

 

I .05(.09)
l STR |
, -.os(.os

 

.61(.O7)'

 

 

 

LSXi l

mu. Model to test Hypothesis 2 with path coefficients. (Beta
weights are given with standard errors in parentheses. * indicates p<
.05. SSN: Social Support Number, SSS=Social Support Satisfaction,
RES=residual term of STR*SSS, LFS=Lifestyle at Time 1,

SXl=Symptoms at Time 1, STR=Stress, SX2=Symptoms at Time 2.)

35
relationship was tested by first identifying a path between SSS and

SX2. Next, the product of the deviation scores of STR and SSS (2(STR-
STR)*(SSS-SSS)) was calculated using SPSS. Then, again using SPSS,

STR and SSS were regressed onto the product mentioned above, and
the residual was obtained. This procedure was undertaken to
maintain the integrity of this new variable as a representation of the
cross between STR and SSS while ensuring that it would not be
correlated with either STR or SSS. Indeed, this was accomplished as
the correlation between the resulting variable, RES, and the product
of the deviation scores was .97 while there was no correlation
between RES and STR or SSS (1:: .00 in both cases). Finally, a path
was set between RES and SX2. If SSS had a moderating effect, RES
should have shown a significant effect on SX2 after the path between
SSS and SX2 was taken into account (Baron & Kenney, 1986). Neither
the path between SSS and SX2 nor that between RES and SX2 was
significant (B=.03, p>.05 in both cases). Therefore, there was no
support for the moderating effect of social support. However, it
should be noted that the mean for Social Support Satisfaction was
140 out of a possible 162, which may indicate a ceiling effect. This
may have attenuated the correlation between RES and SX2.
Therefore, the only significant predictor of SX2 remained SXl (B: .61,
p<.05). SXl also was significantly related to STR (B: .63, p<.05).

The overall Chi-square for the model in Figure 2 was 54.74
(df=5), which gave it a tail probability of .000. This suggests that the
model was a very poor fit to the data. Hypothesis 2 was not

supported.

3 6

Hypothasis 3

For Hypothesis 3, Anxiousness (ANX) and Avoidance (AVO)
were each regressed onto both Social Support Number and Social
Support Satisfaction using SPSS. The mean for ANX and AVO were
31.97 (s.d.: 9.65) and 46.80 (s.d.: 10.32), respectively. Hypothesis 3
was supported by the data. SSN was significantly predicted by both
ANX and AVO (B: -.16, p< .05 and B: -.23, p<.05, respectively), thus
indicating that securely attached persons (people low on ANX and
AVO) tended to have more people in their social networks. In
predicting a subject's SSS, again, both ANX and AVO were significant
(B: -.28, p< .05 and B: -.17, p< .05, respectively), suggesting that
people who are securely attached are more satisfied with their social

networks.

Harms—4.

In order to test hypothesis 4, the full path model was run using
the Least Squares method in PATH. For each endogenous variable a
formula was designed based on the variables in the model which
were purported to be related to that variable. For example, Social
Support Satisfaction was set to be equal to the effects of each of the
two attachment factors. Reliability scores for each of the variables
was used to derive standard errors for each of the path coefficients.
The beta weights, standard errors, and significance levels are shown
in Figure 3.

There were only minor variations between the path coefficients

in the full path model and the beta weights reported earlier in the

 

 

 
   
 
 
 
      
    

 

 

 

 

Has ]
.oa(.07)

 

 

 

 

 

.oa(.07)

 

 

 

-.16(.07)f

.62(.07)'
I LFS

.59(.05)'

 

 

l SX1

 

 

Figural. Path model to test Hypothesis 4 with path coefficients.
(Beta weights are given with standard errors in parentheses. *
indicates p< .05. ANX=Anxiousness, AVO=Avoidance, SSN=Social
Support Number, SSS=Social Support Satisfaction, RES=residual term
of STR*SSS, LFS=Lifestyle at Time 1, SXl=Symptoms at Time 1,
STR=Stress, SX2=Symptoms at Time 2.)

38
analysis of the data for each of the first three hypotheses. SSN was

predicted by ANX (B: -.16, p<.05) and AVO (B: -.22, p<.05), as was
SSS (B: -.29, p<.05, and B: -.17, p<.05, respectively). SSN and SXl
were significant predictors of STR (8: -.12, and B: .59, respectively,
all p<.05). Interestingly, LFS became a significant predictor of STR
when analyzed in this way (B: -.16). Finally, SX2 was only
significantly predicted by SXl. All other paths were non-significant.
The Chi-square for this model was 18.69 (d.f. = 14, p=.177), therefore
this model fit the data somewhat well.

Subsequent analysis using SAS' CALIS produced slightly
different results. While the path coefficients were similar, the
standard errors were slightly smaller. For the path depicted in
Figure 3 (without the residual term), this changed the significance of
the path between LFS and SX2. Additionally, although the goodness
of fit index was .95, which suggests that the model fit the data
reasonably well, the overall Chi-square was significantly different
from that found by PATH. The Chi-square was 51.83 (d.f.=ll, p<.01)
suggesting that the model was a poor fit. The root mean square error
was .12.

It was thought that the difference between the results found
with PATH and CALIS was due to an error in the way PATH
calculates Chi-square. Unfortunately, timing (i.e., deadlines)
prevented a complete rerun of the data on CALIS. Therefore, except
where noted, all of the remaining data analysis was done using
PATH. Because of the problems noted, the conclusions reached in the

rest of this manuscript are to be considered equivocal. (NOTE: If

3 9
anyone would like the analysis done on SAS, please contact the

author.)

Since the originally run path model (Figure 3) was not as
representative of the data as was thought possible, an attempt was
made to find a path model which better fit the data. To this end, all
non-significant paths were deleted. This was done in a step-wise
fashion, one path removed at a time, to test for significant changes in
the model. However, since none of the removed paths individually
caused significant differences between the path models, only the
path model with all of the deleted paths will be presented.

Figure 4 shows the path model after all non-significant paths
were eliminated. Again, ANX and AVO predicted SSS (B: -.29 and B:
-.17, respectively), and SSN (B: - .16 and B: -.22, respectively). STR
was ultimately only significantly predicted by SSN, LFS, and 8X] (B:
-.12, B: -.12, and B: .59, respectively). And SXl was the only
predictor of SX2 (B: .66). The overall Chi-square of this model was
20.49 (d.f.=18, p: .306). Therefore the revised model fits the data
significantly better than the originally tested model.

Although other models were tested (usually created by
changing the direction of a causal arrow) these usually resulted in
only minor changes in both the path coefficients and the significance
of the Chi-square statistic. However, one interesting finding was that
the path between SXl and STR was of the same magnitude regardless
of the direction in which causation was presumed to be.

PATH identified potentially significant paths which were not

tested in this work. One such path was between Social Support

40
,-.29(.07)*

 

 

    
 

 

 

 

 

 

 

 

 

 

 

 

 

l ANX SSS l
-.17(.07)'
-.16(.07)'
l AVO SSN
0'22”"). lambs):
[ STR l SX2 l
-.16(.07)"
.59(.05)' .66(.05)"
l LFS l

 

 

lsml

 

Eigumfi. Modified path model with path coefficients. (Beta weights
are given with standard errors in parentheses. * indicates p< .05.
ANX: Anxiousness, AVO=Avoidance, SSN=Social Support Number,
SSS=Social Support Satisfaction, LFS=Lifestyle at Time 1,
SXl=Symptoms at Time 1, STR=Stress, SX2=Symptoms at Time 2.)

41
Number and Social Support Satisfaction. Another was between

Anxiousness and Stress.

The finding that SSS did not demonstrate a buffering effect was
surprising given previous research. Therefore, an attempt was made
to explore whether SSS had a direct effect. SSS and SSN were
changed in the path model of Figure 2 and a new residual term based
on a product of SSN and STR. However, none of the new paths were
significant and the model fit the data more poorly than the first
model tested.

When the reduced model in Figure 4 (with the LFS-SX2 path
added) was tested, SAS again provided different results than PATH.
The goodness of fit index remained at .95, again suggesting that the
model fit the data reasonably well. However, the Chi-square was
53.13 (d.f.=l3), which also was significant (p<.01). The root mean
square error for this model was .11. As noted earlier, this calls into
question the validity of the data analysis and the resulting

conclusions based on that analysis.

4 2
Discussion

The results suggest that one's level of physical symptoms is
related to the level of stress one experiences concurrently but not to
the physical symptoms one experiences four to six weeks later once
prior symptoms are accounted for. Since causal factors must precede
the factor they are purported to cause, it is unlikely that the stress
experienced at Time 1 caused the symptoms reported at Time 1.
However, since illness presents another demand on one's resources
(e.g. time, energy), it may be that being ill causes stress, and, indeed,
this is supported by other studies. As mentioned earlier, McFarlane,
Norman, Streiner, and Roy (1983) suggested that a person's stress
and health levels may be consistent across time, much like
personality traits. The consistency across models of the beta weights
between stress and illness when measured concurrently supports
this line of reasoning. This suggests that people with a certain level
of symptoms tend to maintain that level over time, and this
consistency of symptoms may increase one's stress level. The design
did not allow the determination of which might come first. A
longitudinal study following people from a very young age may shed
more light on this process.

It may be that the self-report measures of symptoms in this
study were somewhat inaccurate because people under stress may
tend to be more sensitive to unpleasant sensations of their bodies.
More objective measures may be effective in correcting this potential
source of error.

The lack of a significant relationship between stress and

symptoms also may have been an artifact of the time lag between

43
Time 1 and Time 2 in this study (4-5 weeks). It has been suggested

that stress affects a person‘s physical health through its effects on
the immune system (Kiecolt-Glaser et al., 1986). Since the incubation
period of many common illnesses is something less than one week
(Cohen, Tyrrell, & Smith, 1991; Evans, Pitts, & Smith, 1988) and since
most people recover from these illnesses within a period of two to
three weeks, it may be that the time lag in this study was too long to
demonstrate a connection between stress on later illness. Further
studies of this sort should use a one to two week delay between
testing sessions.

People with a healthier lifestyle, including good diet, regular
exercise, and safe sexual habits, tended to suffer from less stress
than those with a less healthy lifestyle, although this relationship
was rather small. It may be that it takes time to exercise and to eat
a more healthy diet. Therefore, it may be that busy students who
are involved in a healthy lifestyle feel added time pressure as well
as reaping the benefits of such good habits. Research devoted to
identifying the pros and cons of healthy habits might be helpful in
revealing such counterbalancing forces.

It was interesting to note that one's lifestyle has little to do
with the amount of symptoms s/he experiences four to six weeks
later, especially in light of the medical community's evidence that
healthy habits tends to reduce health problems. Although the
subjects in this study were consistent about responding to the
lifestyle measure (as can be seen in the test-retest reliability), it may
be that there was an unmeasured social desirability bias. Since it is

well-known that people should exercise, eat well, sleep well, etc.,

44
perhaps the subjects reported that they complied with these

guidelines to "look good" in the eyes of the experimenter. Or perhaps
people reported healthier lifestyles because of a desire to follow that
lifestyle or because of poor recall. Future research might be more
accurate using more objective measures of lifestyle factors, such as
diaries.

The main effects model for social support was partially
supported by this study. Although modest, there was a statistically
significant correlation between the number of people in one's social
support system and the amount of stress one experiences. But, as
reported earlier, that stress was not correlated with subsequent
illness. The size of the relationship between social support and stress
might be increased if more complete measures of the structure of
social support networks were included. These might include
measures of density and integration. There is evidence that the
more integrated one's core network is, the less resistant it is to loss
of one of its members (Hammer, 1983). Therefore, if a person has a
group of friends and family who all know each other, it is more
difficult to deal with the loss of one of that network's members
because everyone in the group is impacted. Also, if one is
experiencing stress due to one or two group members, people with
social networks which are highly integrated might have difficulty
obtaining assistance in dealing with their concerns since all of the
people in the network are very familiar with each other.

The buffering portion of the social support hypothesis was not
supported. This may have been due to the possible ceiling effects

reported earlier in this manuscript. It may also have been due to the

45
nature of the buffering effect. In light of the data on the

physiological effects of denial (Shedler, Mayman, & Manis, 1993),
social support may merely increase the likelihood that a person
experiencing stress will deny the existence or severity of a stressor.
This may increase physical expressions of stress. Such social support
may also interfere with the emotional expression which has been
demonstrated useful in reducing stress (Pennebaker, Kiecolt-Glaser,
& Glaser, 1988). Furthermore, this study did not address the
stressful effects that a social network can sometimes produce (e.g.,
Cohen & Hoberman, 1983; Shaefer, Coyne, & Lazarus, 1981), and,
therefore, it is difficult to determine if a high level of satisfaction
with one's social network is also accompanied by an increase in
responsibilities to the members of the network, and thus an increase
in stress.

This study also supported the idea that people who are anxious
about relationships tend to be less satisfied with their social
networks than those who are not. Avoidant people also tend to be
more dissatisfied with their networks than others, perhaps because
the people in their networks expect closer connections than they are
comfortable with. As a result, it was no surprise to find that people
who tend to avoid relationships and maintain their self-sufficiency
tend to have smaller networks than more outgoing people. The
results also supported the hypothesis that people who are relatively
more anxious about relationships have smaller networks than those
who are relatively less anxious. Both of these findings suggest that
people who are more securely attached tend to have more resources

at their disposal, which tends to reduce stress. If these patterns are

46
consistent throughout the lifespan, there is a strong argument for the

effects of early development of socialization patterns on health later
in life. However, the results of this study do not allow one to make
conclusions about the source of such patterns. If the sources of these
patterns could be identified, it might be possible to intervene to
change such patterns.

It should also be noted that this study did not address how
effectively people of varying levels of anxiousness and avoidance use
their social support networks. If those high on anxiousness or
avoidance tended to make poor use of their networks, it might have
decreased the direct effects of social support on illness and it might
explain the lack of support for the buffering hypothesis.

A couple of notes should be made here. First, at the time this
manuscript was completed, norms for the measures used were not
available. Therefore, it is unknown if the sample used in this study
are representative of the greater population. Second, although the
measures used in this study had adequate reliability, they
necessarily limit the amount of information which can be collected.
If an interview format was used, it would have produced information
more specific to the individual which might have been more
informative from a clinical standpoint. Third, the stresses
experienced by this population are unique. College students face
challenges not commonly encountered by other adults, such as initial
separation from parents, initial encounters with independence, and
classroom demands. Because of these unique stressors, the results of

this study may be difficult to generalize to other populations.

47
Finally, a few notes about the potentially significant paths

which were not tested in this study. One of these was the connection
between social support satisfaction and social support number. It is

possible that people who are more satisfied with their social support
networks would be more likely to retain the people in their networks
and then to augment this steadier base. This would result in peOple

high on satisfaction with their networks having larger networks than E
those less satisfied. Or it may be that people with larger networks 1
are more satisfied with them, perhaps due to the wider range options

this presents them with. If the former is true, it may be that

 

WEAR-u"...

satisfaction with social support affects stress levels indirectly
through the structure of the network. Further study in this area is
warranted to confirm or deny the existence of such a relationship
and to ferret out the nature of it.

The other connection suggested by PATH was between a
person's level anxiousness about relationships and the amount of
stress s/he experiences. It seems likely that people who are anxious
about their interpersonal relationships would experience more stress.
Such a connection would also raise the possibility that it is not the
number of persons in an individual's social support network that
alleviates stress, but rather that one's concern about not having
enough support or the most appropriate kinds of social support
which causes stress. Further research is needed to test these

hypotheses.

LIST OF REFERENCES

48

LIST OFREFERENCES
Ainsworth, M.D.S., Blehar, M.C., Waters, E., & Wall, S. (1978). Patterns
0 .-tt 11‘ 1: ‘ .- Choolica c f th r. 3' i -oti .
Earlbaum Associates: New Jersey.

Andersen, M. (1981). Life changes and health in two drug treatment

modalities. In Schecter, A. J. (Ed.) MW
alguhulism; qulumn 1) Biomadical issuas. New York: Plenum
Press.

Arend, R., Gove, R., & Sroufe, LA. (1979). Continuity of individual
adaptation from infancy to kindergarten: A predictive study of
ego-resiliency and curiosity in pre-schoolers. gala

1.291212121112111. 50. 950-959.

Aro, H., Hanninen, V., & Olavi, O. (1989). Social support, life events
and psychosomatic symptoms among 14- 16-year- old

adolescents. SW, 29(9), 1051- 1056.

Baron, R.M. & Kenny, DA. (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual,
strategic, and statistical considerations. r rs it

and Sugial Psyghulogy, 51, 1173-1182.

Bartholomew, K. & Horowitz, L. M. (1991). Attachment styles among
young adults: A test of a four-category model. Journal at

W 61(2), 226- 244.

Berkman, L.F. & Syme, S. L. (1979). Social networks, host resistance,
and mortality: A nine- year follow- -up study of Alameda County

residents. WW 109(2) 186- 204

Bigby, J. A. (1987). The Nervous System. In Barnes, H. N., Aronson,

M..D & Delbanca T-L- (Eds.) W
121W. New York: Springer-Verlag.

49

Bloom, J. R. (1990). The relationship of social support and health.
Susi al Ssie cuss iu Madisins, 30(5), 635- 637.

Bowlby, J. (1958). The nature of the child's tie to his mother.
lutar r1_1iat Qua] luu utual uf Psysh uaual ysis 3,9 350- 373.

Bowlby, J. (1969). t e t 088' e t (Vol.1). Basic
Books: New York.

Broadhead, W.E., Kaplan, B.H., James, S.A., Wagner, B.H., Schoenbach,
V.J., Grimson, R., Heyden, S., Tibblin, G., & Gehlbach, SH. (1983).
The epidemiologic evidence for a relationship between social

support and health. Amstisau luutual uf Euidsmiolugy, 117(5),
521-537.

Caldwell, R.A. & Reinhart, M.A. (1988). The relationship of
personality to individual differences in the use of type and

source of social support. luurnal uf Sosial aud Qlinical
mm. 6(1). 140-146.

Cannon, W.B. (1929). B i n 'n i h n 11
(2nd ed). New York: Appleton.

Cassel (1976). The contribution of the social environment to host

resistance. Ametisau lourual uf Epiglaiuiulugy, 104(2), 107-
123.

Cobb, S. (1976). Social support as a moderator of life stress.

W 38 300- 314

Cohen, S. (1988). Psychosocial models of the role of social support in

the etiology of physical disease. Health Esychulugy, 7(3), 269-
297.

Cohen, S. & Hoberman, H.M. (1983). Positive events and social
supports as buffers of life change stress. nal f 1'

Susial Esyshulugy, 13(2), 99-125.

Cohen, S. & Matthews, K. A. (1987). Editorial: Social support, type A

behavior, and coronary artery disease. W
Magenta, 49, 325- 330.

50
Cohen, S., Tyrrell, D.A.J., & Smith, A.P. (1991). Psychological stress

and susceptibility to the common cold. New England luurnal of
Medisiue, 325(9), 606-612.

Cohen, S. & Wills, TA. (1985). Stress, social support, and the
buffering hypothesis. Esyc chulugisal Bulletiu, 98(2), 310- 357.

Collins, N. L. & Read, S. J. (1990). Adult attachment, working models,
and relationship quality in dating couples. Juurual of

Estsuualit y and Susi al Esyshulogy, 58, 644- 663.

Cox, D.J., Taylor, A.G., Nowacek, G., Holley-Wilcox, P., Pohl, S.L., &
Guthrow, E. (1984). The relationship between psychological
stress and insulin-dependent diabetic blood glucose control:

Preliminary investigations. HaaltLEsyuquugy, 3(1), 63-75.

Cyr, M.G. & Moulton, A.W. (1987). The gastrointestinal tract and
pancreas. In Barnes, H.N., Aronson, M.D., & Delbanca, T.L. (Eds.),

Alsuhuisui; A guids fut ths pg’maty cars physiciau. New York:
Springer-Verlag.

DeLongis, A., Coyne, J.C., Dakof, G., Folkman, S., & Lazarus, RS. (1982).
Relationship of daily hassles, uplifts, and major life events to

health status. Haalth Esyshulugy, 1(2), 119-136.

Dienstbier, R.A. (1989). Arousal and physiological toughness:
Implications for mental and physical health. Esyshulugisal

Bavisw, 96(1), 84-100.

Dorian, B.J. & Garfinkel, RE. (1987). Stress, immunity and illness - a
review. Psyshulugisal Mauisine, 17(2), 393-407.

Dorian, B.J., Garfinkel, P.E., Keystone, E.C., Gorczynski, R., Garner, D.M.,
Darby, P., & Shore, A. (1985). Occupational stress and

immunity. Psyshusumatis Msglisiue, 47, 77. (As cited in Dorian
& Garfinkel, 1987).

Ende, J. (1987). Nutritional status, cardiovascular, hematologic,
reproductive, and musculoskeletal systems. In Barnes, H. N,

Aronson, M. D, & Delbanca, T. L. (Eds), W
th re h i' . New York: Springer-Verlag.

51

Evans, P.D., Pitts, M.K., & Smith, K. (1988). Minor infection, minor life
events and the four day desirability dip. Juurnal uf

Psyshosutuatis Easearch, 32(4/5), 533-539.

Feeney, J.A. & Noller, P. (1990). Attachment style as a predictor of
adult romantic relationships. 0 r11 1 P rso Ii 11 i

Emmi. 58(2). 281-291.

Fleming, R., Baum, A., Gisriel, M.M., & Gatchel, R]. (1982). Mediating
influences of social support on stress at Three Mile Island.

luurual uf Human Strsss, 8(3), 14-22.

Fontana, A.F., Kerns, R.D., Rosenberg, R.L., & Colonese, KL. (1989).
Support, stress, and recovery from coronary heart disease: A

longitudinal causal model. Haalth Esysholugy, 8(2), 175-193.

Funch, D. P. & Marshall, J. R. (1983). The role of stress, social support
and age in survival from breast cancer. Loam

W, 27(1), 77- 83.

Gannon, L. & Pardie, L. (1989). The importance of chronicity and
controllability of stress in the context of stress-illness

relationships. Juurnal at Behaviural Msdisiua, 12(4), 357-372.

Gannon, L., Banks, J., Shelton, D., & Luchetta, T. (1989). The
mediating effects of psychophysiologial reactivity and recovery
on the relationship between environmental stress and illness.

,loutual uf Esyshusomatis Ressarsh, 33(2), 167-175.

Gore, 8. (1978). The effect of social support in moderating the health

consequences of unemployment. luurual_sf_l-_I_e_a_ltl1_angL_S_Qsial
Sahayiut, 19, 157-165.

Gottlieb, EH. (1985). Social networks and social support: An
overview of research, practice, and policy implications. Haalth

Edusatiun Quartstly, 12(1), 5-22.

Griffin, D.W. & Bartholomew, K. (in press). The metaphysics of
measurement: The case of adult attachment. In K.

Bartholomew & D. P. Perlman (Eds), W

WW3; London:
Jessica Kingsley.

52
Hammer, M. (1983). 'Core' and 'extended' social networks in relation

to health and illness. Sucial Ssience in Medisiue, 17(7), 405-
41 1.

Havel, R.J. (1989). Dietary and hereditary factors in coronary heart

disease. In Carroll, K.K. (Rd), W (pp. 19-
32). Montreal & Kingston: McGill-Queen's University Press.

Hazen, C. & Shaver, P. (1987). Romantic love conceptualized as an
attachment process. r ' n i

12mm, 52. 511-524.

Holahan, C.K., Holahan, C.J., & Belk, SS. (1984). Adjustment in aging:
The roles of life stress, hassles, and self-efficacy. Health

Wit. 3(4). 315-328.

House, J.S., Robbins, C., & Metzer, H.L. (1982). The association of
social relationships and activities with mortality: Prospective
evidence from the Tecumseh Community health study.

Ametisan laughal at Euidsmiulugy, 116(1), 123-140.

Hunter, J.E. & Hamilton, M.A. (1992). PATH. An unpublished
statistical package designed for path analysis.

Jacobs, T.J. & Charles, E. (1980). Life events and the occurrence of
cancer in children. Psychosuiuatis Madisina, 42(1), 11-24.

Jemmott, J.B. & Locke, SE. (1984). Psychosocial factors, immunologic
mediation, and human susceptibility to infectious diseases: How

much do know? Psyshglugisal Bullstiu, 95(1), 78-108.

Kanner, A.D., Coyne, J.C., Schaefer, C., & Lazarus, RS. (1981).
Comparison of two modes of stress measurement: Daily hassles

and uplifts versus major life events. luurual uf Bahaviural
Magicina, 4(1), 1-39.

Kaplan, N. M. (1989). Diet and hypertension. In Carroll, K.K. (Ed.),

MW (pp. 93-102) Montreal & Kingston:
McGill- Queen's University Press.

Katch, F..I & McArdle, WD. (1993). W
W. Philadelphia/London: Lea & Febiger.

53

Keast, D. & Morton, A.R. (1992). Long-term exercise and immune

functions. In Watson, R.R. & Eisinger, M. (Eds.), W4
diseasa (pp. 89-120), Boca Raton, FL: CRC Press. _

Kiecolt-Glaser, J .K., Glaser, R., Strain, E.C., Stout, J.C., Tarr, K.L.,
Holliday, J.E., & Speicher, CE. (1986). Modulation of cellular

immunity in medical students. Wow;
9(1), 5-21.

Kobak. R. (1986). MW.

Unpublished manuscript, University of ‘Denver. As cited in
Kobak, R.R. & Sceery, A. (1988). Attachment in late
adolescence: Working models, affect regulation, and

representations of self and others. Wm, 59, 135-
146.

Kobak, R.R. & Sceery, A. (1988). Attachment in late adolescence:
Working models, affect regulation, and representations of self

and others. W. 59, 135-146.

Kobasa, S..,C Maddi, S..,R & Puccetti, M.C. (1982). Personality and
exercise as buffers in the stress- illness relationship. Juumal_g_f

BehaxieLaLMedicine 5(4) 391 -.404

Kotler, T. & Omodei, M. (1988). Attachment and emotional health: A
life span approach. W, 41(8), 619-640.

Lazarus. RNA & Folkman S (1989) W

W. Consulting Psychologists Press: Palo
Alto, CA.

Lee, R. N. F., Graydon, J. E., & Ross, E. (1991). Effects of psychological
well- -being, physical status, and social support on oxygen-
dependent COPD patients' level of functioning. 8mm

W101 l4 323- 328

Lutkenhaus, P., Grossmann, K.E., & Grossmann, K. (1985). Infant-
mother attachment at twelve months and style of interaction

with a stranger at the age of three years. W
56, 1538- 1542.

54

McFarlane, A..H, Norman, G..,R Streiner, D..,L & Roy, R..G (1983). The
process of social stress: Stable, reciprocal, and mediating

relationships WWW 24(June).
160 173.

Minter, R.E. & Kimball,C.P. (1979). Life events and illness onset: A
review. Lsyshdsdmatids, 19, 334-339.

Monroe, S. M. (1983). Major and minor life events as predictors of
psychological distress: Further issues and findings. Jdurua1__Qf

W. 6(2), 189- 205.

Moulton, A..W & Cyr, M..G (1987). The liver. In Barnes, H..,N

Aronson, M.,D. & Delbanca, T..L (Eds), Alcuhui 1stu; A guide et'ut
W. New York: Springer-Verlag.

Mrazek, D. A., Casey, B. & Anderson, I. (1987). Insecure attachment
in severely asthmatic preschool children: Is it a risk factor?
1-. 1 1‘ .11 -1‘ 111 1 h1_ 11 .11'C1

Esichiatu 26(4) 516- 520.

Norbeck, J.S. & Tilden, VP. (1983). Life stress, social support, and
emotional disequalibrium in complications of pregnancy: A
prospective, multivariate study. LaumalJLHaaltLaMial
Bahayiut, 24, 30-46.

Orth-Gomér, K. & Undén, A. (1987). The measurement of social
support in population surveys. ° ' 11 ' i he,
24(1), 83-94.

Pennebaker, J. W., Kiecolt-Glaser, J. K, & Glaser, R.(1988). Disclosure
of traumas and immune function: Health implications for

psychotherapy MSW.
56(2), 239 245.

Rabkin, J.G. & Struening, EL. (1976). Life events, stress, and illness.
m, 194 (Dec.), 1013-1020.

Sarafino, EP. (1990). WWW
Wiley _& Sons: New York.

 

55

Sarason, I.G., Levine, H.M., Basham, R.B., & Sarason, BR. (1983).
Assessing social support: The Social Support Questionnaire.
Per nal' a S il P hl , 44(1), 127-139.

Sarason, I.G, Sarason, B. R, Potter, E. H. & Antoni, M. H. (1985). Life
events, social support, and illness. ati e
47(2), 156- 163.

Schaefer, C., Coyne, J.C., & Lazarus, RS. (1981). The health-related

functions of social support. luurnal at Behaviural Madicins,
4(4), 381-405.

Selye, H (1956). flfhs stress uf lifs. McGraw-Hill: New York.

Shedler, J., Mayman, M., & Manis, M. (1993). The illusion of mental
health. Amerisah Esycholugist, 48(11), 1117-1129.

Sheldon, A.E.R. & West, M. (1990). Attachment pathology and low
social skills in avoidant personality disorder: An exploratory

study. Canadiau Journal at Esyshiatty, 35, 596-599.

Siegrist, J., Siegrist, K., & Weber, 1. (1986). Sociological concepts in
the etiology of chronic disease: The case of ischemic heart

disease. Susial Ssisuce in Medisina, 22(2), 247-253.

Simpson, LA. (1990). Influence of attachment styles on romantic

relationships. luumal uf Persuuality and Susial Esyshulugy,
59(5), 971-980.

Simpson, J. A, Rholes, W. S., & Nelligan, J. S. (1992). Support seeking
and support giving within couples in an anxiety- -provoking

situation: The role of attachment styles. luur ual of Bar rsuhality
and Susial Esysholo og,y 62(3), 434- 446.

Sroufe, LA. (1983). Infant-caregiver attachment and patterns of
adaptation in pre-school: The roots of maladatation and

competence. In M. Perlmutter (Ed), W
W (Vol. 16, pp. 41- 81). Hillsdale, NJ: Earlbaum.

Stokes, J. P. & Wilson, D. G. (1984). The Inventory of Socially
Supportive Behaviors: Dimensionality, prediction, and gender

differences. ri na ni P l ,
12(1), 53- 69.

56

Stone, A.A., Bruce, R., & Neale, J.M. (1987). Changes in daily event
frequency precede episodes of physical symptoms. Juurnal uf

Human Strass, 13, 70-74.

Sugimura, T., Wakabayashi, K., Nagao, M., & Ohgaki, H. (1989).
Mutagens and carcinogens formed during cooking. In Carroll,

K.K. (Ed.), Dist, hugitiun, Q health (pp. 207-228). Montreal &
Kingston: McGill-Queen's University Press.

Thoits, RA. (1982). Conceptual, methodological, and theoretical
problems in studying social support as a buffer against life

stress. luutnal uf Health and Sosial Behaviur, 23(June), 145-
159.

VanderPlate, C., Aral, S.O., & Magder, L. (1988). The relationship
among genital herpes simplex virus, stress, and social support.

Health Psyshology, 7(2), 159-168.

Verde, T.J. (1992). Short-term exercise and immune function. In

Watson, R.R. & Eisinger, M. (Eds.), Exersiss and dissase, (pp. 71-
88). Boca Raton, Fl.: CRC Press.

Wienberger, M., Hiner, S.L., & Tierney, W.M. (1987). In support of
hassles as a measure of stress in predicting health outcomes.

luutnal at Bahavioral Msdisiua, 10(1), 19-31.

Wills, T.A., & Langner, TS. (1980). Supportive functions of
interpersonal relationships. In S. Cohen & S.L. Syme (Eds.),

MW (pp. 61-820). New York: Academic

Press.

Wolever, T.M.S. & Jenkins, D.J.A. (1989). Diet and diabetes. In

Carroll, K.K. (Ed.). WM (pp- 103-120)-
Montreal & Kingston: McGill-Queen's University Press.

Zarski, 1.]. (1984). Hassles and health: A replication. Health
Psyshulogy, 3(3), 243-251.

 

"llllllllllllllllllllls