Content uploaded by Margaret E Kemeny
Author content
All content in this area was uploaded by Margaret E Kemeny
Content may be subject to copyright.
Journal of Personality and Social Psychology
1998,
Vol. 74, No. 6, 1646-1655
Copyright 1998 by the American Psychological Association, Inc.
OO22-3514/98/S3.OO
Optimism Is Associated With Mood, Coping, and Immune Change
in Response to Stress
Suzanne C. Segerstrom, Shelley E. Taylor, Margaret E. Kemeny, and John L. Fahey
University of California, Los Angeles
This study explored prospectively the effects of dispositional and situational optimism on mood (N
= 90) and immune changes (N = 50) among law students in their first semester of study. Optimism
was associated with better mood, higher numbers of helper T cells, and higher natural killer cell
cytotoxicity. Avoidance coping partially accounted for the relationship between optimism and mood.
Among the immune parameters, mood partially accounted for the optimism—helper T cell relationship,
and perceived stress partially accounted for the optimism-cytotoxicity relationship. Individual differ-
ences in expectancies, appraisals, and mood may be important in understanding psychological and
immune responses to stress.
Recent years have witnessed substantial progress in under-
standing the contribution of psychosocial factors to physical
health. One such factor, optimism, or the expectation of positive
outcomes, has been tied to better physical health (Scheier et al.,
1989) and more successful coping with health challenges
(Carver et al., 1993; Stanton & Snider, 1993). However, the
routes by which optimism might be associated with better health
have not received systematic investigation. One plausible route
is through effects on the immune system. Optimists cope differ-
ently with stressors, experience less negative mood, and may
have more adaptive health behaviors, all of which could lead
to better immune status. The present investigation examined
optimism in the context of a major stressor, namely, the first
year of law school, specifically examining relationships among
optimism, mood, and immune changes. Coping and health be-
haviors were also examined as potential routes for these effects.
Suzanne C. Segerstrom and Shelley E. Taylor, Department of Psychol-
ogy, University of California, Los Angeles; Margaret E. Kemeny, Depart-
ment of Psychology and Department of Psychiatry and Biobehavioral
Sciences, University of California, Los Angeles; John L. Fahey, Depart-
ment of Microbiology and Immunology, University of California, Los
Angeles.
This article is based on a dissertation by Suzanne C. Segerstrom
submitted to the University of California, Los Angeles, in partial fulfill-
ment of the requirements for the PhD in psychology. It was supported
by the National Institute of Mental Health (Grants MH10841, MH15750,
and MH00820), the UCLA Norman Cousins Program in Psychoneu-
roimmunology, and the National Institute of Allergy and Infectious Dis-
eases (Grant N01A172631).
We thank Dean Barbara Varat of the University of California, Los
Angeles (UCLA), School of Law; Susan Plaeger, Susan Stehn, and
Pablo Villanueza of the Center for Interdisciplinary Research in Immu-
nology and Disease; and research assistant Thomas deHardt. We also
thank Constance Hammen, Hector Myers, and Annette Stanton for their
helpful comments on drafts of this article.
Correspondence concerning this article should be addressed to Su-
zanne C. Segerstrom, who is now at the Department of Psychology,
University of Kentucky, Lexington, Kentucky, 40506-0044. Electronic
mail may be sent to scsegeO@pop.uky.edu.
Optimism has been shown to mitigate the effects of stressors
on psychological functioning. Dispositional optimists (who hold
generalized positive outcome expectancies) have shown less
mood disturbance in response to a number of different stressors,
including adaptation to college (Aspinwall & Taylor, 1992;
Scheier & Carver, 1992), breast cancer biopsy (Stanton &
Snider, 1993), and breast cancer surgery (Carver et al., 1993).
These findings may be attributed to optimists' belief that dis-
crepancies between their goals and their current attainment will
be resolved, minimizing defeat-related moods such as shame,
depression, and anger (Carver & Scheier, 1985).
Optimism has also been associated with better physical
health. Dispositional optimists reported better physical health
(Scheier & Carver, 1992), showed fewer signs of infarct during
coronary artery bypass surgery (CABG), and reported better
quality of life after surgery (Fitzgerald, Tennen, Affleck, & Pran-
sky, 1993; Scheier et al., 1989). F. Cohen et al. (1989) found
that dispositional optimists had more T lymphocyte immune
cells than pessimists in response to stressors lasting less than 1
week, though the opposite was true in response to stressors
lasting more than 1 week. Situational optimism about health
outcomes with respect to HIV has been associated with slower
immune decline (Kemeny, Reed, Taylor, Visscher, & Fahey,
1998),
later symptom onset (Reed, Kemeny, Taylor, & Visscher,
in press), and longer survival time in AIDS (Reed, Kemeny,
Taylor, Wang, & Visscher, 1994).
Mediational Paths Between Optimism
and Immune Change
Because optimism is reliably associated with less negative
mood, mood is a plausible first pathway by which optimism
could be associated with immune changes under stress. Clinical
states,
primarily major depression, but also generalized anxiety
and posttraumatic stress, have been associated with fewer circu-
lating lymphocytes and poorer lymphocyte function (Herbert &
Cohen, 1993; Ironson et al., 1997; La Via et al., 1996). In addi-
tion, subclinical mood disturbance, changes in daily mood, and
experimentally induced mood have correlated with lymphocyte
1646
OPTIMISM AND IMMUNE CHANGE
1647
number, function, or both (Futterman, Kemeny, Shapiro, & Fa-
hey, 1994; Stone, Cox, Valdimarsdottir,
Jandorf,
& Neale, 1987;
Stone et al., 1994; Zorrilla, Redei, & DeRubeis, 1994).
Another means by which optimism could result in immune
differences is coping. Dispositional optimists make less use of
avoidance strategies such as denial and giving up, which has
accounted for mood differences between optimists and pessi-
mists (Aspinwall & Taylor, 1992; Carver et al., 1993; Scheier,
Weintraub, & Carver, 1986; Stanton & Snider, 1993; Taylor et
al.,
1992). Avoidance coping, a passive coping style, and denial
have been associated, mainly cross-sectionally, with worse im-
mune status in both healthy and clinical samples (Futterman,
Wellisch, Zighelboim, Luna-Raines, & Weiner, 1996; Goodkin,
Blaney, et al., 1992; Goodkin, Fuchs, Feaster, Leeka, & Rishel,
1992;
Ironson et al., 1994; Kedem, Bartoov, Mikulincer, &
Shkolnik, 1992; Kemeny, 1991).
Health behavior constitutes a third pathway by which opti-
mism may be associated with immune changes under stress.
Scheier and Carver (1987) suggested that optimists have more
positive health habits as a function of their generally more adap-
tive coping style. Furthermore, as they use less avoidance to
cope with stressors, they may use less alcohol as an avoidance
strategy; they may sleep better because they have less depression
and anxiety. Health behavior such as alcohol use and sleep are
known to affect immune parameters (Kiecolt-Glaser & Glaser,
1988).
The Present Study
The first goal of the present study was to examine the degree
to which optimism, both dispositional and situational, was asso-
ciated prospectively with changes in mood disturbance and im-
mune parameters during a stressor. A second goal was to assess
whether mood, coping strategies, and health behaviors account
for these effects.
The first year of law school has been reported, both anecdot-
ally and in research, to be extremely stressful (Clark & Rieker,
1986;
Heins, Fahey, & Leiden, 1984; Turow, 1977). We expected
that, because it is stressful, the first two months of law school
would be associated with an increase in negative mood and
changes in the immune system, specifically, in the number and
function of lymphocytes, immune cells found in peripheral
blood. Among lymphocyte subsets, natural killer (NK) cells and
T cells are particularly sensitive to naturalistic psychological
stressors in young, healthy adults. These two lymphocyte subsets
have been found to decrease in number in medical school stu-
dents taking examinations (Glaser et al., 1985; Glaser, Rice,
Speicher, Stout, & Kiecolt-Glaser, 1986). Decreases in NK cell
cytotoxicity have also been observed in that context (Glaser et
al.,
1986).
We predicted that situational and dispositional optimism
would mitigate these effects. We expected that optimists would
have less negative mood than pessimists. In terms of the immune
system, we expected that optimists would also have higher num-
bers of lymphocytes and higher NK cell cytotoxicity than pessi-
mists under stress. These differences might be due to optimists'
better mood, more adaptive coping, or healthier behavior.
Method
Participants
In June 1994, the University of California, Los Angeles (UCLA)
School of Law mailed recruitment packets for the study to all successful
applicants intending to attend, a group of approximately 375 students.
Materials included a study description, informed consent form and infor-
mation about the rights of human participants in medical experiments.
Participants were instructed, if interested in participating in the study,
to return a signed informed consent form, telephone contact information,
and screening questionnaire in the return envelope provided.
One hundred five students
(31%
of the 337 students who matriculated)
returned screening materials in time to participate in the study. Potential
participants were excluded from the study if examination of screening
materials revealed any of the following: previous psychiatric hospitaliza-
tion, use of psychiatric medication in the previous 3 months, or severe
psychological distress (such that the person could not function in her
or his usual work or activities for 2 weeks or more) in the previous 3
months. One student reported psychological distress and was excluded
from the study.
Of the 104 eligible participants, 99 returned questionnaires at Time
1,
and 94 returned questionnaires at Time 2.' Following data collection,
four participants' data were excluded from analysis: One participant
reported taking antidepressant medication during the study period, and
3 reported serious life stressors during the study period, which might
have confounded results (e.g., 1 participant was divorcing).
The final sample for the study, then, was 90 first-year law students.
Mean age of the sample was 23.9 years (range = 20-37). The sample
was about equally divided between men (51.1%) and women (48.9%).
The racial-ethnic makeup of the sample was as follows: 54.4% of the
sample was White, 8.9% was Hispanic-Chicano-Latino, 15.5% was
Asian American (including Pacific Islander), and 11.1% was African
American. Ten percent of the sample either indicated that they were of
mixed race-ethnicity or gave responses from which we could not deter-
mine their race-ethnicity. The majority of the sample (90%) was single
and childless.
Participants had to meet strict eligibility criteria to have immune
measures taken. These criteria were intended to ensure that any findings
represented changes in a healthy and normally functioning immune sys-
tem. This necessarily reduced the number of participants for analyses
including immune measures. However, power analyses had indicated that
stress-immune effects would be detected with a sample size of 50.
Exclusion criteria were moderate to severe anxiety about venipuncture,
past or current immunologically mediated disease (e.g., rheumatoid ar-
thritis),
possible current infection, anemia, habitual alcohol use (e.g.,
two or more drinks daily), anesthesia in the previous 3 months, major
medical illness (e.g., diabetes), or use of medication or drugs that could
affect the immune system. Fifty-eight participants completed question-
naires and had blood drawn at Time 1, and 53 completed questionnaires
and had blood drawn at Time 2. Following data collection, 3 additional
participants' immune data were excluded from analysis because they
reported taking prescription medication that might affect the immune
system.
The final sample for immune study was 50 participants. The demo-
graphic characteristics of these participants did not significantly differ
from those of the other participants. Furthermore, the two groups did
not significantly differ on other Time
1
measures, including dispositional
optimism, situational optimism, and mood.
' Participants were lost to follow-up because they left law school or
failed to return questionnaires. In addition, anemia and failed venipunc-
ture resulted in some participants not giving blood at Time 2.
1648
SEGERSTROM, TAYLOR, KEMENY, AND FAHEY
Procedure
Data were collected in two waves. Time 1 data were collected during
the 2 weeks preceding law school orientation and the first day of classes.
Time 2 data were collected during 2 weeks at midsemester (Weeks 8
and 9 of a 16-week semester).
Participants having blood drawn were contacted by telephone and
scheduled for venipuncture, which occurred between 7:00 a.m. and 9:00
a.m. at both time points to control for circadian changes. Participants
were asked not to drink alcohol or caffeinated beverages, smoke, or
exercise in the morning before venipuncture. Participants came to a
room at the law school where a phlebotomist drew 50 ml of blood
from an antecubital vein into sterile, preservative-free, evacuated tubes
(Vacutainer Systems, Becton Dickinson, Rutherford, NJ). Participants
were given a questionnaire packet containing psychological measures
with instructions to complete the measures the same day. These partici-
pants were paid $30 at each time point.
Participants not eligible for blood draw were contacted by phone and
informed of the procedures of the study. Each of these students received,
in his or her law school mailbox, questionnaires that contained the
psychological measures with instructions to complete the questionnaire
during the data collection period (e.g., the 2 weeks before classes
started) and return it to the experimenter. These participants were paid
$10 at each time point for their participation. All materials were coded
to protect participants' confidentiality.
Questionnaire Measures
Dispositional optimism: Life Orientation Test (LOT). The LOT
(Scheier & Carver, 1985) measures dispositional optimism, which is
defined as generalized positive outcome expectancies. Four items are
positively phrased ("In uncertain times, I usually expect the best"),
and four are negatively phrased ("If something can go wrong for me,
it will"). An additional four items are fillers. Respondents indicate their
agreement with each item on a 5-point scale ranging from strongly agree
(1) to strongly disagree (5). The LOT has acceptable psychometric
properties and discriminant validity with respect to related concepts
such as locus of control and helplessness. The LOT was administered
at Time 1.
Situational optimism. A 10-item scale was designed for use in this
study. This scale measures three aspects of specific optimism based on
previous research in optimism about HIV (Reed et al., in press) and
adapted for the first semester of law school. These aspects were perceived
risk of failure
("It's
unlikely that I will fail"), optimistic bias ("I will
be less successful than most of my classmates"), and confident emotions
("I feel confident when I think about it"). Five items were phrased
positively and five, negatively. Respondents indicated their agreement
with each statement on a 5-point scale ranging from strongly agree (1)
to strongly disagree (5).
The situational optimism scale was administered at Time 1 and Time
2.
The internal reliability of the scale was .86 at Time 1 (as measured
by coefficient alpha) and .91 at Time 2. The test-retest correlation was
.66.
2
The correlation between dispositional and situational opti-
mism at Time 1 was .30, suggesting that these two constructs were
discriminable.
Coping. The Coping Operations Preference Enquiry (COPE; Carver,
Scheier, & Weintraub, 1989) is a coping inventory that participants
completed at Time 2 for each of two stressors identified in pretesting.
3
The COPE has meta-factors, including Problem Solving, Mental Accom-
modation, and Avoidance. These factors were found in two separate
validation samples (Carver et al., 1989). Problem Solving includes active
coping, planning, and suppression of competing activities; Mental Ac-
commodation includes acceptance and positive reinterpretation and
growth; and Avoidance includes denial, mental disengagement, and be-
havioral disengagement. A ninth first-order factor, focus on and venting
of emotions, was included, as emotional approach has been conceptual-
ized as a mental accommodation strategy (Stanton, Danoff-Burg, Cam-
eron, & Ellis, 1994).
Mood:
Profile of Mood States (POMS). The POMS (McNair,
Lorr, & Droppleman, 1971) is a measure of mood state over the previous
week. Respondents rate how much they have been feeling each of 65
different moods on a 5-point scale ranging from not at all (0) to ex-
tremely (4). The scale yields Total Mood Disturbance, which comprises
subscales measuring tension-anxiety, depression-dejection, anger—hos-
tility, fatigue-inertia, vigor-activity, and confusion-bewilderment. The
POMS has high internal consistency (.74 to .91 for the subscales) and
good validity. The POMS was administered at Time 1 and Time 2.
Health behavior. Participants were asked about their health behav-
iors over the 7 days preceding questionnaire administration. Use of
caffeine, nicotine, alcohol, and drugs was assessed, as well as days of
aerobic and anaerobic exercise and average hours of sleep nightly.
A number of these health behaviors were not normally distributed,
had outliers (values lying more than three standard deviations from the
mean),
or both. Values for outliers were set equal to the next highest
value in the distribution. This change was necessary for 2 participants'
caffeine use reports at Time 1 and for
1
participant's alcohol use reports
at Time 1 and Time 2. Positively skewed distributions—caffeine use,
number of drinks over a week, and days of aerobic exercise over a
week—were Iog
10
-transformed, improving normality. Finally, cigarette
smoking was dichotomized into a smoker-nonsmoker variable, because
only 9 participants reported smoking at either time point.
Demographic and personal characteristics. Participants reported
their age, sex, race-ethnicity, marital status, and how many children
they had living with them (if any). They were also asked to report their
law school aptitude test (LSAT) results in terms of a raw score and
percentile.
4
Stressors. Participants were asked to describe in a free-response
format "other extremely stressful experiences in your life lately" and
to rate such stressors on a 7-point Likert-type scale that ranged from 1
(not at all stressful) to 7 (the most stressful thing I have ever
experienced).
2
Although .66 is low by conventional standards of reliability, the
measure in question would be expected to change over time as partici-
pants gained experience with the situation. The test-retest reliability
suggests moderate stability, all that would be expected of a situational
measure. As the main objective of the study was to predict mood and
immune change prospectively, reported analyses were limited to use of
the Time 1 measure. Analyses using the Time 2 measure revealed slightly
higher correlations with other Time 2 questionnaire measures (mood,
coping), a result that might be attributed to shared measurement vari-
ance.
Time 2 situational optimism was not related to any immune
measures.
3
In the year prior to the study, questionnaires listing stressors associ-
ated with law school (generated from Clark & Rieker, 1986; Heins et
al.,
1984) were distributed to the first-year law students 8 weeks into
their first semester. Response rate was approximately
18%.
Two stressors,
understanding legal material and lack of positive feedback, were rated
by students as uniformly stressful (M = 4.28 on a 7-point scale) and
were selected for the purposes of the present study. Pretesting also
indicated that stressors impacted male and female law students equally
(Segerstrom, 1996). Analyses using separate stressfulness ratings and
coping scores from the two stressors did not result in different results.
Therefore, results collapsing across the two stressors are reported.
Stressfulness ratings correlated .38; problem solving, .47; mental accom-
modation, .61; and avoidance, .83.
4
Controlling for LSAT scores did not affect the results of the study;
therefore, this variable is not discussed further.
OPTIMISM
AND
IMMUNE CHANGE
1649
Immune Measures
Immune measures included number
of
cells
in
four lymphocyte
sub-
sets:
CD4
+
cells (helper
T),
CD3
+
CD8
+
cells (cytotoxic
T),
CD19
+
cells (B), andCD3~CD16+56
+
cells (NK). Natural killer cell cytotox-
icity (NKCC)
was
also measured. Following collection, samples were
kept
at
room temperature
for no
more than
2.5 hr and
were then taken
to the Clinical Immunology Research Laboratory
in
the Center
for
Inter-
disciplinary Research in Immunology and Disease at UCLA
for
immuno-
logic analysis. Laboratory personnel were unaware
of
the hypotheses
of
the study.
All immune measures were examined
for
outliers (i.e., values more
than three standard deviations from
the
group mean).
One
value
for
absolute number
of
CD3
+
CD8
+
cells
and
two values
for
absolute num-
ber
of
CD19
+
cells were dropped from analysis because they were
outliers. Examination
of
immune parameters
for
the
3
participants with
outlying values suggested that
the
outliers were anomalous, because
all
other immune values were well within normal limits.
Lymphocyte subset analysis. Fifty microliters
of
whole blood that
had been collected
in
EDTA anticoagulant
was
incubated with
0.01
ml
of
fluorescein isothiocyanate (FITC-), phycoerythrin (PE-),
and
peridinin chlorophyll protein (PerCP-) conjugated murine antihuman
monoclonal antibodies (MAb)
for 15 min at
4° C. After incubation,
the
cells were washed once
and
tested immediately.
The
samples were
ana-
lyzed with
a
FACScan flow cytometer (Becton Dickinson,
San
Jose,
CA) equipped with
a
15-W argon laser. List mode data were stored
and
analyzed with FACScan research software (Becton Dickinson, San Jose,
CA).
Lymphocytes were identified
by
gating
on
forward (low-angle)
and 90° (wide-angle) light scatter parameters
and
verified using
a
com-
bination
of
anti-CD45/CD4 monoclonal antibodies. Isotype controls
were used
to
evaluate nonspecific binding
and to
position cursors. MAb
were purchased from Becton Dickinson Immunocytometry System
Inc.
(San Jose, CA). Three-color analysis was conducted with the following
combinations
of
FITC,
PE, and
PerCP-labeled
MAb:
CD19/CD62L/
CD4
and
CD3/CD56+16/CD8.
White blood cell
and
differential counts
on
whole blood were
per-
formed
on a
Coulter MD16 instrument (Hialeah,
FL) to
obtain total
lymphocyte counts, which were used
to
calculate absolute numbers
of
the cells within the lymphocyte subsets. The total number
of
lymphocytes
was multiplied
by
each subset percentage obtained
by
flow cytometry
to obtain
the
absolute numbers
of
lymphocytes with specific phenotypic
antigens.
Natural killer cell cytotoxicity (NKCC). Heparinized whole periph-
eral blood was layered
on
Ficoll-Hypaque gradient (Histopaque; specific
gravity,
1.077;
Sigma,
St.
Louis, MO). Gradients were centrifuged
at
800 X gravity
for
10 min. The cell interface, which contained the mono-
nuclear cell fraction,
was
harvested
by
aspiration,
and the
cells were
washed twice with Hank's balanced saline solution (GIBCO, Life Tech-
nologies,
Inc.,
Gaithersburg,
MD). A
viable cell count
was
done
by
trypan blue dye exclusion. The effector cell preparation was resuspended
in RPMI-1640 with
5%
newborn calf serum
at a
concentration
of 5 x
10
6
lymphocytes/ml.
Cytotoxicity was measured
in a
standard 3-hr chromium release assay
with K562 cells
as the
target cells.
The
target cells were labeled with
51
Cr as
sodium chromate. Triplicate aliquots (0.1
ml) of
the target cell
suspension were added
to
the effector cell suspensions
at
three effector-
target ratios
(50:1,
25:1, 12.5:1)
in
96-well microliter plates.
Use of
several effector-target ratios
is
customary
to
provide
a
comprehensive
evaluation
of the
functioning
of
this cell system. Spontaneous release
was determined in wells that contained only target cells; maximal release
was determined
in
wells with target cells
in
media that contained deter-
gent (Triton-100).
The
plates were incubated
for 3 hr in a 5% CO
2
incubator
at 37° C.
They were centrifuged,
and
100-^1 aliquots were
removed
to
count with
a
gamma counter (Gamma-Tm 1193,
Tm
Ana-
lytic,
Brandon,
FL) the
amount
of
sodium chromate
(
5l
Cr)
released.
Percent lysis was calculated
by
using
the
following formula: (cpm sam-
ple
-
cpm spontaneous)/(cpm maximum release
- cpm
spontaneous),
where
cpm
equals counts
per
min. Lytic units were
not
calculated
be-
cause percent lysis
was too low in
many cases.
Percent lysis (NKCC)
was
divided
by
percent
NK
cells
to
yield
adjusted NKCC (aNKCC). Effector-target cell ratios were calculated
using lymphocytes,
not
only
NK
cells,
as
effector cells. However, target
cells
are
susceptible
to
killing only from
NK
cells. When the proportion
of
NK
cells
in the
effector
mix
drops,
the
effective ratio
of
cytotoxic
cells
to
target cells also drops, potentially yielding misleading results
(Kiecolt-Glaser
&
Glaser, 1991). Expressing NKCC
as a
ratio between
percent lysis
and
percent
NK
cells
in the
effector
mix
corrects
for
this
problem
(cf.
Naliboff
et al.,
1995).
Results
Mood
and
Immune Changes Over Time
Mood and immune parameters were expected to change from
Time 1 to Time 2, reflecting the stressful nature of the first
semester of law school. From Time 1 to Time 2, mood distur-
bance increased from a mean of 1.24 (SD = 0.51) to a mean
of 1.55 (SD = 0.58), r(88) = 6.42, p <
.0001.
5
Repeated
measures multivariate analysis of variance (MANOV\) on the
four enumerative immune measures showed a trend toward a
Cell Type x Time interaction, F(3, 43) = 2.52, p < .07). In
individual comparisons, there was a significant change only in
the number of NK cells, which decreased from a mean of 336
cells at Time 1 (SD = 178) to a mean of 281 cells (SD = 147),
t(49) =
-2.55,
p < .01. Repeated measures MANOVA on the
three aNKCC ratios indicated a significant increase in aNKCC,
F(l, 48) = 5.02,p < .03, which varied somewhat across ratios,
F(2,47) =
3.31,
p < .05. Post hoc analyses showed that change
at the 12.5:1 and 25:1 ratios was significant, t(49) = 2.35 and
2.39, respectively, ps < .03, whereas change at the 50:1 ratio
only approached significance, f(49) = 1.93, p < .06.
6
Effects
of
Optimism
on
Mood
and
Immune Change
Optimism was predicted to be associated with better mood
and higher lymphocyte subset numbers and function at Time 2.
As predicted, both dispositional and situational optimism were
associated with less mood disturbance at Time 2, both before
and after controlling for Time 1 mood
(
see Table 1). Situational
optimism had a slightly stronger relationship to negative mood.
Table 2 shows the correlations between dispositional and situ-
ational optimism and the immune measures at Time 2, partialing
the corresponding Time 1 immune measure. Optimism, and in
particular situational optimism, was related to higher lympho-
cyte subset numbers and function. Dispositional optimism was
positively associated, though not significantly, with higher num-
bers of cytotoxic T (CD3
+
CD8
+
) cells. Situational optimism
was similarly related to number of cytotoxic T cells. In addition,
situational optimism was significantly positively correlated with
number of helper T (CD4
+
) cells and with aNKCC at the 12.5:1
5
Two participants were missing Time
1
mood data.
6
For purposes
of
comparison with other research,
we
conducted
an
exploratory analysis using unadjusted NKCC. There
was no
change
in
unadjusted NKCC from Time
1 to
Time
2, F(l, 48) =
0.07,
p > .05.
1650
SEGERSTROM, TAYLOR, KEMENY, AND FAHEY
Table 1
Correlations Between Optimism and Mood
Variable
1
1.
Dispositional optimism
2.
Situational optimism
3.
Time 2 mood
4.
Time 1 mood
Bivariate
.30**
-.33**
-.25*
-.39**
-.28**
.70**
—
Partialing Time 1 mood
1.
Dispositional optimism —
2.
Situational optimism .23*
3.
Time 2 mood -.23*
-.28*
*p < .05. **p < .01.
and 25:1 effector-target ratios; the correlation with aNKCC at
the 50:1 ratio just failed to reach significance (p < .06).
Correlates of Optimism
The relationships among optimism, mood, and immune
changes might be due to differences in stress appraisals, to
differences in coping, or to variations in health behaviors. As a
first
step,
we examined correlations between optimism and these
variables.
Coping strategies and stress appraisal were examined first.
To
confirm the factor structure of the COPE scales, we submitted
coping scores to confirmatory factor analysis (CFA) using EQS
software (Bentler & Wu, 1995) As has been true in prior re-
search (Carver et al., 1989), the model had a Problem-Solving,
a Mental Accommodation, and an Avoidance factor, which were
allowed to covary. Focus on and venting of emotions loaded on
the Avoidance factor, rather than on Mental Accommodation.
This may be due to the fact that this strategy is confounded
with distress (Stanton et al., 1994), and therefore might also be
expected to load with other coping strategies associated with
distress during chronic stressors (i.e., avoidance; Roth & Cohen,
1986).
The three-factor model was an acceptable fit,
7
x
2
(24, N =
90) = 25.02, p > .05 (Bentler-Bonett nonnormed fit index
[BBNNFI] =
1.00).
There was a significant correlation be-
tween Problem Solving and Mental Accommodation (r = .82,
p < .0001). There were smaller correlations between Mental
Accommodation and Avoidance (r =
.21,
p > .05) and between
Problem Solving and Avoidance (r = —
.14,
p > .05).
Some data have suggested a two-factor, second-order struc-
ture of the COPE (e.g., Carver et al., 1993) that separates avoid-
ant and nonavoidant coping scores. In this case, the two-factor
structure was not as good a fit to the data, x
2
(26,
AT
= 90) =
39.58,
p < .05; BBNNFI = .93. Comparison of the models
verified that the three-factor solution was a significantly better
fit,
x
2
(2,
N = 90) = 14.56, p < .05.
Correlations between optimism, perceived stress, and coping
were calculated. Time 1 mood disturbance was partialed out, as
concurrent distress might be related to optimism (Smith, Pope,
Rhodewalt, & Poulton, 1989). Dispositional optimism and situa-
tional optimism were significantly associated with less avoid-
ance coping (dispositional optimism, r = -.2\,p < .05; situa-
tional optimism, r = -.27, p < .05). Situational optimism was
also significantly associated with less perceived stress (r =
-.28,/? < .05).
Finally, health behaviors were examined. However, only one
health behavior correlated with optimism: Hours of sleep at
Time 2 (partialing Time 1 sleep) correlated with situational
optimism, an effect that was marginally significant (r = .20, p
< .10).
Roles of Coping, Stress, and Health Behavior in
Optimism Effects
Both types of optimism were associated with more avoidance
coping, and situational optimism was also associated with higher
perceived stress and fewer hours of sleep. These relationships,
therefore, could account for mood and immune changes associ-
ated with optimism. Furthermore, mood differences between
optimists and pessimists could account for correlations between
situational optimism and lymphocyte subset numbers and
function.
The first analyses examined effects of coping and perceived
stress on the association between optimism and negative mood.
As optimism was significantly correlated with both avoidance
coping and mood, and avoidance coping was itself significantly
correlated with mood (after partialing baseline mood, r = .43,
p < .0001), coping could mediate the relationship between
optimism and mood (Baron & Kenny, 1986; Holmbeck, 1997).
To test this possibility, correlations between optimism and mood
were repeated, partialing baseline mood and avoidance coping.
8
For dispositional optimism, the partial correlation between opti-
mism and mood dropped from —.23 (p < .05) to —.15 (p <
.16) after controlling for coping, suggesting partial mediation
of the effect. However, the difference between the two coeffi-
cients was not significant (z
diff
= 0.77, p > .05). The magnitude
of this z score indicates that the decrease after controlling for
coping was modest and should be interpreted with caution. Cop-
ing appeared to account for no more than part of the effect of
dispositional optimism on mood. Turning to situational opti-
mism, the correlation between optimism and mood was reduced
from -.28 (p < .05) to -.19 (p < .08) after partialing avoid-
ance coping. Again, this difference was modest (z
diff
= 0.89, p
> .05). Overall, the results indicate that a portion of the effect
of optimism, both dispositional and situational, on mood was
due to less avoidance coping by optimists; however, this portion
was modest and did not account for the entire effect.
Situational optimism was also associated with perceived
stress,
which itself significantly correlated with negative mood
7
Fit was measured by chi square and fit index. When a model fits
the obtained data, the chi-square probability exceeds alpha (in this case,
.05).
The BBNNFI Index is a measure of fit that takes into account the
degrees of freedom of the model. Values above .90 are desirable (Bentler,
1995).
8
Structural equation models of Time 1 and Time 2 mood, optimism,
and coping were attempted; however, the stability of the mood measure
led to overdetermined models. A simpler method of analysis was there-
fore chosen.
OPTIMISM AND IMMUNE CHANGE
1651
Table 2
Correlations Between Optimism and Immune Parameters
Optimism
Dispositional
Situational
CD4
+
.01
.35*
CD8
+
.25t
•24t
CD19
+
.15
.08
CD16 + 56
+
-.01
.01
12.5:1
.00
.28*
aNKCC ratio
25:1
.06
.31*
50:1
.12
.27t
Note. Correlations are with Time 2 immune measures, partialing out the corresponding Time 1 measure.
CD4
+
= helper T cells; CD8
+
= cytotoxic T cells (also CD3
+
); CD19
+
= B cells; CD16 + 56
+
= natural
killer cells (also CD3~); aNKCC = adjusted natural killer cell cytotoxicity.
t Marginally significant atp < .10 *p < .05.
(r = .21, p < .05). Perceived stress therefore could mediate
the relationship between situational optimism and mood. The
correlation between situational optimism and mood was there-
fore repeated, partialing Time 1 mood and perceived stress. The
correlation was reduced from -.28 (p < .05) to -.19 (p <
.08).
The magnitude of this difference {nm = 0.89, p > .05)
was similar to that obtained with coping, suggesting that per-
ceived stress could account for part of the relationship between
situational optimism and mood.
The second set of analyses examined mediators of the sig-
nificant relationships between situational optimism and number
of helper T cells and NKCC. We recalculated the correlations
between situational optimism and these immune parameters,
partialing out the corresponding Time 1 immune measure and
each potential mediator: mood, perceived stress, avoidance cop-
ing, and hours of sleep (see Table 3).
The relationship between optimism and number of helper T
cells changed very little after controlling for perceived stress,
avoidance coping, and sleep. However, a modest reduction re-
sulted from controlling for mood disturbance, which itself corre-
lated with number of helper T cells (r = —.47, p < .001).
The correlation between situational optimism and helper T cells
dropped from .35 to .21 (zam = 0.95, p > .05). These findings
suggest that the relationship between situational optimism and
number of helper T cells was explained, in part, by changes in
mood disturbance.
Table 3
Partial Correlations Between Situational Optimism
and Immune Parameters
Correlation
Initial
Partialing
Mood
a
Stress
Avoidance
Sleep'
CD4
+
.35*
.21
.33*
.31*
.36*
12.5:1
.28*
.26t
.18
.27t
.27t
aNKCC ratio
25:1
.31*
.29*
.25t
.31*
.30*
50:1
.27t
.25t
.25t
.30*
.26t
Note. Correlations are with Time 2 immune measures, partialing out
the corresponding Time 1 immune measure. CD4
+
= helper T cells;
aNKCC = adjusted natural killer cell cytotoxicity.
a
Mood and sleep at Time 1 and Time 2 were partialed.
t Marginally significant at p < .10. * p < .05.
A different result occurred with aNKCC. The magnitude of
correlations changed very slightly, if at all, after controlling for
mood, avoidance coping, and sleep. The only substantial de-
crease in the correlation between situational optimism and
aNKCC occurred after controlling for perceived stress. This
decrease was most noticeable at the 12.5:1 ratio, at which the
correlation between optimism and aNKCC dropped from .28 to
.18,
though even at this ratio the magnitude of the change was
rather small (z
diff
= 0.68, p > .05). The decrease was even less
marked at the 25 : 1 ratio and was very slight at the 50:1 ratio.
This can be attributed to the magnitude of correlation between
perceived stress and aNKCC, which was highest at the 12.5:1
ratio (r = -.23, p < .11) and lower at the 25:1 (r = -.18, p
< .23) and 50:1 (r = -.08, p < .57) ratios.
Discussion
Optimism has been associated with better psychological and
physical adjustment to stressful events. Results of the present
study suggest that optimism may also be associated with im-
mune change during stressful circumstances. Specifically, stu-
dents in their first semester of law school who scored high on
situational optimism by endorsing an optimistic bias (Taylor &
Brown, 1988; Weinstein, 1980), expectations for success, and
confident emotions in regard to their first semester of law school
tended to have higher lymphocyte subset number and function.
These results add to an emerging body of literature to suggest
that appraisal of stressful events relates to concomitant immune
changes (Ironson et al., 1997; Kiecolt-Glaser et al., 1987).
The immune changes that varied with optimism are generally
thought to be beneficial ones. Optimists had more helper T
cells,
an essential immunoregulatory cell that mediates immune
reactions to infection. Similarly, among HIV-seropositive gay
men, situational optimism about future health was associated
with higher numbers of helper T cells (Kemeny et al., 1998);
this immune parameter has been found to be prognostic of health
changes in HIV. Optimists in the present study also had higher
NKCC. NKCC is thought to be important in mediating immunity
against viral infection and some types of cancers. The changes
observed here in healthy participants, although substantial, were
within normal limits, and such changes may not be clinically
relevant. Whether this magnitude of immune change would af-
fect reactions to a health challenge remains a question for future
study.
The increase in NKCC seen in this study from before starting
1652 SEGERSTROM, TAYLOR, KEMENY, AND FAHEY
law school to midsemester contrasts with a stress-related NKCC
decrease that has been found, for example, in medical students
during examination periods (Glaser et al., 1986). There are two
possible explanations for this contrast. First, NKCC decrease
in other studies might have been an effect of decreased percent-
age of NK cells in the assay (Kiecolt-Glaser & Glaser, 1988).
Although this study adjusted for this effect, most previous re-
search has not done so. Second, optimism and appraisals may
offer a clue. The present participants were studied two months
before their first examination period. Within this time, optimists
showed an increase in NKCC and number of helper T cells,
whereas pessimists stayed roughly the same or decreased in
these immune parameters (see Table 4). Optimists may have
seen this period as a challenging, positive experience rather than
as threatening, and thus showed increases in immune parameters
rather than the decreases that are often associated with taxing
or threatening stressors.
Situational optimism was a stronger predictor of mood than
dispositional optimism and predicted immune changes where
dispositional optimism did not. Several theories of behavior and
affect predict that situation-specific cognitions predict in that
situation better than trait constructs (e.g., Ajzen & Fishbein,
1977;
Bandura, 1977; Lazarus,
1991;
Weiner, 1986). In research
on psychosocial concomitants of HIV infection, HIV-specific
optimism predicted behavior, mood, immune status, and health
changes better than dispositional optimism (Kemeny et al.,
1998;
Reed et al., in press, 1994; Taylor et al., 1992). The
present study adds additional evidence to the observation that
situation-specific appraisals may predict reactions to specific
situations better than more general measures and provides con-
verging evidence that these effects extend to immune changes
as well. Moreover, the present results add credence to the more
general methodological and measurement concern regarding the
need to match the level at which cognitions are assessed to the
context in which they occur, whether general or specific.
Mediators of Optimism—Immune Relationships
Three variables were evaluated for their potential role as me-
diators of the relationship between optimism and immune
change: mood, coping strategies, and health habits. Mood has
Table 4
Immune Values Associated With Situational
Optimism or Pessimism
Time 1 Time 2
Immune parameter/group" M SD M SD Change
CD4
+
Situational optimists 833 309 943 226 +13%
Situational pessimists 873 380 849 240 -3%
aNKCC (25:1 ratio)
Situational optimists 3.02 1.57 4.29 2.01 +42%
Situational pessimists 3.15 1.23 3.44 1.90 +9%
Note. CD4
+
= helper T cells; aNKCC = adjusted natural killer cell
cytotoxicity.
a
Group was based on a median split.
predicted neuroendocrine and immune changes in a number of
studies (Futterman et al., 1994; Herbert & Cohen, 1993; Ironson
et al., 1990; Ironson et al., 1997; Linn, Linn, & Jensen, 1981;
Stone et al., 1987; Stone et al., 1994; Zorrilla et al., 1994), and
it accounted for part of the relationship between situational
optimism and number of helper T cells in this study. However,
the relationship between situational optimism and NKCC was
not accounted for by mood. It may be that the current mood
measurement was too blunt to reveal a relationship between
optimism, mood, and NKCC. Future research might explore
alternative means of measuring mood, perhaps including mea-
sures specific to the stressor, more sensitive to daily mood
changes (Stone et al., 1987; Stone et al., 1994), or less sensitive
to self-report biases (e.g., Stroop tasks).
Alternatively, appraisal may have been associated with NKCC
in a manner unrelated to mood. Maier, Watkins, and Fleshner
(1994) reasoned that "thoughts ought to be capable of altering
immunity" (p. 1009) to the degree that they have come to be
associated with aversive events. Negative appraisals or expectan-
cies might lead people to find aversive or threatening meaning
in their circumstances. In this case, situational optimism and
NKCC were related to some degree through perceived stress, a
variable that likely reflects the degree to which students saw law
school as aversive. Other research has suggested that cognitive
factors may lead to immune changes independent of changes in
mood. In three investigations, negative expectations about health
and negative self-views predicted the rate of decline of helper T
lymphocyte number and lymphocyte function in HIV infection,
independent of negative mood (Kemeny et al., 1998; Kemeny &
Dean, 1995; Segerstrom, Taylor, Kemeny, Reed, & Visscher,
1996).
In addition, experimental priming of anxio- and de-
pressogenic cognitions has been associated with lower NKCC
(although negative emotions were also increased by priming;
Strauman, Lemieux, & Coe, 1993).
Although avoidance coping has been associated with lower
lymphocyte subset numbers and function in other studies, it did
not mediate the relationship between optimism and immune
change in this study. However, avoidance coping did account
for some of the mood effects of optimism, consistent with other
research (Aspinwall & Taylor, 1992; Carver et al., 1993; Stan-
ton & Snider, 1993; Taylor et al., 1992). There are several ways
in which avoidance coping might have led to more mood distur-
bance in pessimistic students. First, students who avoided threat-
ening stimuli associated with law school might not have been
studying as well or as often. Falling behind in the voluminous
reading associated with the first semester of law school could,
in turn, have increased distress. Second, the behavioral disen-
gagement component of avoidance may have acted as a measure
of helplessness, which has been associated with affective distur-
bance, particularly depression (Abramson, Seligman, & Teas-
dale,
1978). Third, efforts to avoid distressing thoughts or situa-
tions associated with law school may have had a paradoxical
effect: In the long run, rather than helping people minimize
distress, avoidance may increase distress, possibly by increasing
intrusive thoughts (Wegner, 1989) or preventing full processing
of threatening stimuli (Foa & Kozak, 1986; Rodriguez & Craske,
1993).
Some investigators have attributed the positive effects of opti-
mism to its inverse relationship with neuroticism, especially
OPTIMISM AND IMMUNE CHANGE
1653
as concerns self-reported health (Smith et al., 1989; Williams,
1992).
In reply, Scheier and Carver (1992) pointed out that
neuroticism is a multifaceted construct that may include ele-
ments of optimism and pessimism; optimism would therefore
be expected to correlate with neuroticism. In the present investi-
gation, optimism predicted coping and negative mood above and
beyond initial levels of negative mood, which were measured
concurrently with optimism and controlled in all analyses. How-
ever, mood measures are only moderately correlated with nega-
tive affectivity measures (Watson & Clark, 1984). Therefore,
this explanation cannot be definitively ruled out in terms of
the relationship between optimism and mood. The relationship
between situational optimism and immune changes, however,
offers evidence for the discriminant validity of optimism with
respect to neuroticism. Although neuroticism is strongly associ-
ated with symptom reporting, it is less clear that neuroticism is
associated with actual physical changes (S. Cohen et al., 1995;
Watson & Pennebaker, 1989).
Limitations
Some limitations of the present study that are due to the
sample should be considered. First, this was a relatively small
sample that represented approximately one third of the first-
year law school class. The sample was further reduced for some
analyses because of inclusion criteria necessary to ensure valid
interpretation of immune results. Self-selection may have oc-