Acute Threat to the Social Self: Shame, Social Self-esteem, and

Cortisol Activity
TARA L. GRUENEWALD, PHD, MARGARET E. KEMENY, PHD, NAJIB AZIZ, MD, AND JOHN L. FAHEY, MD
Objective: Our Social Self Preservation Theory asserts that situations which threaten the “social self” (ie, one’s social value or
standing) elicit increased feelings of low social worth (eg, shame), decrements in social self-esteem, and increases in cortisol, a
hormone released by the hypothalamic-pituitary-adrenal axis. To test our theoretical premise, cognitive, emotional, and physio-
logical responses to the performance of laboratory stressor tasks were compared in participants who performed these tasks in the
presence or absence of social-self threat. Methods: Pre- and poststressor emotion, self-esteem, heart rate, blood pressure, and
salivary cortisol were compared in 81 participants randomly assigned to complete speech and mental arithmetic stress tasks with
social evaluation present (n ⫽ 41) or absent (n ⫽ 40). Results: As hypothesized, participants in the social evaluation condition
exhibited greater increases in shame and greater decrements in social self-esteem. Other psychological states (eg, anxiety,
performance self-esteem) did not show differential changes as a function of the social context. Salivary cortisol increased in social
evaluation condition participants but did not increase in participants who performed the same tasks in the absence of social
evaluation. Cortisol increases were greater in participants who experienced greater increases in shame and greater decreases in
social self-esteem under social-self threat. Conclusion: Threat to the social self is an important elicitor of shame experience,
decreases in social self-esteem and cortisol increases under demanding performance conditions. Cortisol changes may be
specifically tied to the experience of emotions and cognitions reflecting low self-worth in this context. Key words: stress, cortisol,
shame, emotion, self-esteem.

DBP ⫽ diastolic blood pressure; SBP ⫽ systolic blood pressure;
HR ⫽ heart rate; SOC-EVAL ⫽ social evaluation condition; NON-
EVAL ⫽ non-evaluation condition; ABS ⫽ Affect Balance Scale;
SSGS ⫽ State Shame and Guilt Scale
INTRODUCTION
tinent to a primary human goal: that of achieving and main-
taining a positive “social self.”
Humans are argued to be social animals that possess fun-
damental motivations to belong and be accepted by others
(2– 4), and the status of one’s social self is argued to be an
important determinant of the distribution of social and phys-

W e propose that threats to the social self, or situations

which threaten to demean one’s social image or stand-
ing, engender a specific set of psychological and physiological
reactions. Our position, which we have labeled the Social Self
Preservation Theory (1), is that these events elicit increases in
feelings of low social worth (eg, shame, humiliation), decre-
ments in social self-esteem, and increases in cortisol, a hor-
mone of the hypothalamic-pituitary-adrenal (HPA) system.
We define threats to the social self as situations that contain
the potential to devalue one’s social self by calling into
question abilities, competencies, or traits on which a positive
social image is based (eg, evaluative situations, such as giving
a presentation at work or school), or situations characterized
by potential or explicit rejection. We assert that such situations
are provocative because they contain social information per-
From the University of California, Los Angeles, Los Angeles, CA (T.L.G.,
N.A., J.L.F.); and the University of California, San Francisco, San Francisco,
CA (M.E.K.).
Address correspondence and reprint requests to Tara L. Gruenewald, De-
partment of Medicine/Geriatrics, UCLA David Geffen School of Medicine,
10945 LeConte Ave, Suite 2339, Los Angeles, CA, 90095-1687. E-mail:
tlgruen@ucla.edu or tgruenewald@mednet.ucla.edu
Received for publication December 23, 2003; revision received June 12,
2004.
This research was supported by a National Research Service Award
(MH12703) from the National Institute of Mental Health, an American
Psychological Association Dissertation Research Award, a Society for the
Psychological Study of Social Issues Grants-In-Aid Award, and a UCLA
Graduate Division Dissertation Research Award supporting the first author.
Tara L. Gruenewald is currently in the Department of Medicine/Geriatrics,
Najib Aziz is in the Department of Pathology and Laboratory Medicine, and
John L. Fahey is in the Department of Microbiology and Immunology, at the
UCLA David Geffen School of Medicine, University of California, Los
Angeles. Margaret E. Kemeny is in the Health Psychology Program at the
University of California, San Francisco.
DOI: 10.1097/01.psy.0000143639.61693.ef
ical resources (5,6). Thus, concerns with achieving or main-
taining a positive social self may be an important human
motive that is supported by specific psychological and phys-
iological systems that provide signaling (ie, “the social self is
under threat”) and mobilizing (ie, initiation of psychobiolog-
ical activity to help deal with the threat) functions.
Psychological Responses to Social-Self Threats
Numerous social psychological, evolutionary, and emotion
theories converge to suggest that situations which threaten the
social self should elicit the experience of a specific class of
emotions, those that can broadly be referred to as feelings of
low social worth: shame, humiliation, embarrassment, etc
(7–11). A number of theorists have asserted that shame is the
prototypical emotion experienced in response to devaluations
of the social self (6 – 8,12–14). Phenomenological studies also
support this hypothesis, as people have reported that previous
experience of shame was characterized by a sense of social
isolation, feeling small and inferior, and a pressure to hide
from others (15,16). To date, however, there is little experi-
mental research on this topic; most of the previous research
employed imagined scenarios thought to elicit shame emo-
tions or recall of past situations in which these emotions were
experienced (15–18).
Acute threats to the social self are also likely to affect one’s
thoughts and opinions of oneself. This may be especially true
for the social component of state self-esteem: thoughts regard-
ing the positive or negative state of the social self (eg, feeling
inferior, feeling as if one has made a poor social impression).
Leary and associates (19,20) have demonstrated that imagined
or actual social rejection in the laboratory leads to lowered
overall state self-esteem and more negative ratings on a num-

Psychosomatic Medicine 66:915–924 (2004) 915

0033-3174/04/6606-0915
Copyright © 2004 by the American Psychosomatic Society

ber of emotions (eg, anxious, upset, happy, calm, ashamed)
and self-relevant adjectives (eg, self-confident, superior, ef-
fective, smart). However, their postrejection assessments have
not clearly distinguished shame from other emotional re-
sponses or effects on social vs. nonsocial (eg, performance,
appearance) dimensions of self-esteem, which are goals of the
present investigation.
Physiological Responses to Social-Self Threats
One physiological system that we believe may be provoked
by threats to the social self is the HPA axis, a hormonal system
that regulates metabolism and energy production and is
thought to be centrally involved in regulating physiological
responses to stress (21–24). A large body of human and
animal literature indicates increased levels of HPA hormones
(eg, corticotrophin releasing hormone, adrenocorticotrophic
hormone, and cortisol/corticosterone) under physically and
psychologically threatening conditions. Narrative reviews
suggest that certain characteristics of situations may be more
likely to activate the HPA system, including situations char-
acterized by novelty, uncontrollability, unpredictability, and
ego involvement (24 –27).
Threat to the social self as an important elicitor of the HPA
axis has received less attention in the stress literature, but
social challenges, especially those marked by negative evalu-
ations from others and loss of status, have been identified as
important elicitors of HPA stress responses (see 27). Acute
and chronic social rejection, important forms of social-self
threat, have been found to correlate with cortisol levels.
Stroud and associates (28) demonstrated that a social rejection
experience in the laboratory led to short-term increases in
salivary cortisol, especially in females. Similar associations
are also found in more naturalistic environments; children
socially rejected at school have higher basal cortisol levels
than their accepted or even neglected peers (29). Negative,
unsupportive social environments are argued to contribute to
alterations in HPA functioning (30); however, threat to the
social self as an important component of this association has
received less attention. Nonhuman animal research also indi-
cates that levels of and changes in HPA hormones are sensi-
tive to social threats, including subordinate social status or the
threat of status loss (important nonhuman animal forms of
threat to the “social self;” eg, 31–34).
Threat to the social self may also play an important role in
short-term cortisol elevations witnessed in response to acute
stressor experience in the laboratory. Public speaking tasks
and activities that tax cognitive resources, such as the combi-
nation of speech and mental arithmetic tasks included in the
well-known Trier Social Stress Test (TSST; 35), have been
shown to elicit transient elevations in cortisol in the labora-
tory. However, the specific characteristics of acute stressor
paradigms that engage the HPA axis are unclear. In a meta-
analytic review of 208 laboratory acute stress studies, Dick-
erson and Kemeny (36) demonstrated that stressor paradigms
characterized by social-evaluative threat, such as stressful
situations in which an evaluative audience was present or in
916
T. L. GRUENEWALD et al.
which the participant was the target of a negative social
comparison, led to greater cortisol reactivity as compared with
paradigms in which social-evaluative threat was absent or
minimized. This meta-analytic review provides strong support
for the hypothesis that factors that threaten the social self
within laboratory stressor paradigms may be potent elicitors of
HPA responses. To date, however, we know of no acute stress
investigation that has compared cortisol reactivity to a stressor
task that manipulated the presence or absence of social-eval-
uative threat. The present investigation provides an opportu-
nity to experimentally validate the conclusions drawn by the
meta-analysis and our theoretical premise that the HPA sys-
tem will be activated by threat to the social self.
The Present Investigation
In the present investigation, we tested the hypothesis that a
situation that had the potential to threaten the social self would
be more likely to elicit increases in shame, decrements in
social self-esteem, and increases in cortisol, than a situation in
which threat to the social self was minimized. To do this, we
manipulated participants’ performance of “stressful” activities
(speech and math tasks) so that half of the individuals per-
formed these activities in the presence of an unfriendly, eval-
uative audience (“social-evaluative condition,” SOC-EVAL),
to create a situation in which participants’ social selves might
be threatened, whereas the other half performed the demand-
ing tasks in private (“nonevaluative condition,” NON-EVAL)
so that potential threat to the social self was minimized.
Engaging in the tasks in either condition required cognitive
resources (participants had to prepare their speech, solve dif-
ficult math problems), demanded active performance (oral
delivery of speech, solve math problems under time pressure),
and was characterized by uncontrollability (math problems
became increasingly difficult and impossible to solve),1 char-
acteristics of situations typically thought to be “stressful”
(25–28). The manipulated factor that differed between the two
stress conditions was the presence or absence of direct social-
evaluative threat.
We were interested in whether increases in cortisol would
be more likely to occur in the social-evaluative as compared
with the nonevaluative stressor; however, we also included
measures of heart rate (HR) and blood pressure to assess
changes in arousal that we believed would occur in both stress
conditions.
METHODS
Participants
Eighty-one undergraduate college students (34 male, 47 female) partici-
pated in the experiment. Participants ranged in age from 17 to 22 years (M ⫽
18.67). The majority of participants identified their ethnicity as Asian-Amer-
ican (44.9%) or Caucasian (32.1%), with smaller numbers of participants
identifying as Black/African American (1.3%), Latino/a, Chicano/a (6.4%),
Filipino/a (3.8%) or “Other” (11.5%). Participants for this experiment were
drawn from a larger investigation examining relationships between social
1
In this context, uncontrollability refers to the fact that behavior will have
little, if any, effect on the outcome of the situation.
Psychosomatic Medicine 66:915–924 (2004)
SOCIAL-SELF THREAT
Figure 1. Timeline of session activities and psychological and physiological measurements.
status and mental health in college dormitories. The results of this larger study
are not reported here.
All participants were screened for and excluded from participation if they
possessed health conditions or engaged in behaviors that might affect the
biological measures assessed in this study, including a number of health
conditions (eg, Cushing’s disease, high blood pressure) and use of prescrip-
tion (eg, oral contraceptives) and nonprescription (eg, marijuana, regular
tobacco use) drugs. Participants were asked not to engage in strenuous
exercise, drink alcohol, or smoke on the day of their appointment,2 and to not
consume dairy products, caffeine, or eat 1 hour before the session.
Participants were paid $20 or given psychology course experimental
credit, whichever they preferred, for their participation in the study.
Measures
Questionnaire Measures
Measures of emotion and self-esteem were administered before and after
the stress tasks (see below). For the prestress administration participants were
asked to answer each question in response to how they felt during the last few
minutes, and for the poststress administration how they felt during last ten
minutes, including during the speech and math tasks.
The Affects Balance Scale (ABS; 37), a 40-item measure of positive and
negative state emotion experience (1 ⫽ never to 5 ⫽ always), was used to
measure state emotion. Three additional emotion items (humiliated, self-
conscious, embarrassed) were added to more carefully assess shame-related
emotions. Pre- and poststress mean scores for the anxiety (nervous, timid,
tense, anxious, afraid; internal reliability coefficient Cronbach’s ␣pre ⫽ 0.80,
␣post ⫽ 0.80) and shame (ashamed, humiliated, self-conscious, and embar-
rassed; ␣pre ⫽ 0.86, ␣post ⫽ 0.86) subscales were examined in analyses. The
shame subscale of the State Shame and Guilt Scale (SSGS; 38) was used as
a second measure of pre- and poststress shame experience. Participants
responded to five statements (eg, “I want to sink into the floor and disappear”)
on a five-point scale (1 ⫽ not feeling this way at all to 5 ⫽ feeling this way
very strongly; ␣pre ⫽ 0.69, ␣post ⫽ 0.83).
The social and performance subscales of the State Self-Esteem Scale (39)
were also administered before and after the performance tasks. Respondents
rated their agreement with 14 statements on a five-point scale (1 ⫽ not at all
to 5 ⫽ extremely; performance ␣pre ⫽ 0.85, ␣post ⫽ 0.88; social ␣pre ⫽ 0.86,
␣post ⫽ 0.91).
A task rating questionnaire consisting of three questions assessing feelings
of difficulty, threat, and challenge felt during the activities (seven-point scale,
1 ⫽ extremely poor/not at all true to 7 ⫽ extremely well/very true) was
administered after the performance tasks. Participants in the social stress
condition were also asked to rate how much they liked the panel of evaluators,
how much they thought the panelists liked and accepted them, how effective
they believed they were in convincing panelists that they should get the job
during the speech portion of the activity, and how well they thought the
panelists would rate their performance overall (see below for protocol de-
scription), using a 7-point scale (1 ⫽ not at all/very poor to 7 ⫽ very
much/very excellent).
Physiological Measures
Measures of HR, systolic blood pressure (SBP) and diastolic blood pres-
sure (DBP) were assessed with a Critikon automatic sphygmomanometer
2
Although habitual smokers were excluded from participation, participants
were directed not to smoke on the day of their appointment in the event that
an occasional smoker might smoke on the day of the appointment. All
participants indicated that they did not smoke on the day of their appointment.
Psychosomatic Medicine 66:915–924 (2004)
(Dinamap Model 8146, Critikon Corporation, Tampa, FL). Measurements
were taken oscillometrically by placing an occluding cuff over the brachial
artery of the participant’s nondominant arm. Readings were taken every 2
minutes during specific phases of the experimental session (see below) and
averaged to produce a single value for each phase.
Salivary cortisol was assessed at four points throughout the experimental
session (see below) with the Salivette device (Sartstedt, Rommelsdorf, Ger-
many). This device is essentially a cotton roll which participants were asked
to place under their tongue for 3 minutes. The saturated roll was placed into
a sealed, plastic container, refrigerated for no longer than 16 hours, and then
centrifuged to extract saliva. Samples were frozen at ⫺80°C until assayed.
Cortisol was measured using a high sensitivity salivary cortisol immunoassay
kit (Salimetrics, State College, PA; sensitivity ⬍0.007 ␮g/dl). Values are
reported in ␮g/dl.
Design and Procedure
Participants reported individually for a 90-minute laboratory session
scheduled between the hours of 2:30 and 7:00 PM to control for diurnal
variations in cortisol. A timeline of major study activities is provided in
Figure 1. During the first 20 minutes, participants provided informed consent,
completed prestress questionnaires, and rested quietly. Baseline HR and BP
readings were taken during the last 10 minutes of the rest period, and the first
saliva sample was taken immediately after (“baseline” measures).
Participants were then given tape-recorded instructions in which they
were instructed that they would be asked to perform a set of challenging and
demanding laboratory tasks, which would include having to give a 5-minute
speech and then perform a 5-minute mental arithmetic task on a computer.
The tasks represent a modified version of the TSST (35). In the speech,
participants were asked to explain why they would be good candidates for a
job as a member of a focus group in which they would have to work
effectively with other students to devise solutions to problems typically faced
by college students. A procedure borrowed from Pruessner et al. (40) was
used for the arithmetic task. During the 5-minute task, subjects were presented
with addition, subtraction, multiplication, and division problems and were
given 5 seconds to select the correct answer from a range of 1-digit choices.
The task became increasingly difficult and eventually reached a point in
which participants could not solve the problems in the allotted time. An
annoying noise played during display of the problems, and wrong answers or
failure to answer in the time allotted was followed by a loud buzz.
In the SOC-EVAL condition of the study, participants were informed that
they would perform the activities in front of a panel of evaluators who would
judge their performance, while participants in the nonevaluative (NON-EVAL)
condition were informed that they would perform these activities while alone
in the room.
HR and BP readings were taken while the participants listened to task
instructions (“task instruction” measure). The participants then had 10 min-
utes to mentally prepare their speech. A second saliva sample was taken at the
end of this preparation period (“stress anticipation” measure).
The participant was then taken to a second area of the room and asked to
stand in front of a microphone. In the SOC-EVAL condition, the panel of
evaluators was already seated behind a table facing the microphone. The panel
of evaluators consisted of two individuals of approximately the same age as
the participant who were trained to respond in a nonaccepting and critical
manner (eg, nonsmiling, stony faces, writing comments on their clipboard).
The participant was told that the panelists were present to judge his/her
performance and ability to communicate ideas successfully in a social situa-
tion. The experimenter introduced the participant to the panelists, turned on an
917

audiotape recorder,3 demonstrated how the experimenter would monitor
sound from the other room to ensure they spoke for the entire 5 minutes,4 and
then asked the participant to wait until she left the room to begin his/her
speech. The experimenter was not present during the speech so that she would
not act as an additional source of social evaluation. When an alarm signaled
the end of the speech segment, the experimenter reentered the room and
moved the participant to the computer terminal in order to begin the math
task. Once the experimenter left the room, the participant then completed the
math activity on the computer within the view of the panel. HR and BP were
monitored during the speech and math tasks.
In the NON-EVAL stressor condition, the participant was asked to per-
form these same tasks while alone in the room. The participant was asked to
give his or her speech facing the table that the evaluators sat at in the
SOC-EVAL condition and to complete the same computerized math task;
however, no evaluators were present. As in the SOC-EVAL condition, the
speech was audiotaped and sound was monitored.
After the stressor tasks, participants in both conditions were taken to
another area of the room where they were asked to sit and complete poststress
questionnaire measures. Five minutes after the cessation of the stressor,
participants were asked to provide another saliva sample (“poststress” mea-
sure). After the saliva sample, participants were asked to rest quietly for 5
minutes. HR and BP were monitored during this resting period, and the last
saliva sample was taken at the end of the rest period (“resting recovery”
measures). Finally, participants were fully debriefed and paid or awarded
experiment credit.
Statistical Analysis
Possible baseline differences between SOC-EVAL and NON-EVAL con-
dition participants on all dependent variables were examined with t tests to
ensure success of random assignment. Possible gender differences in levels
and stress reactivity of cardiovascular and salivary measures, and possible
time of day of laboratory appointment effects for salivary cortisol were
assessed with repeated-measures ANOVAs.
Repeated-measures ANOVAs with stress condition (SOC-EVAL vs.
NON-EVAL) as a between-subjects factor and time of measurement of
psychological and physiological measures as within-subjects variables were
performed to assess hypotheses of differential psychobiological stress re-
sponses in SOC-EVAL and NON-EVAL stress condition participants. With-
in-subjects psychological variables were pre- and poststress levels of (a)
anxiety (ABS) and shame (ABS and SSGS) emotions and (b) performance
and social self-esteem. Physiological within-subjects variables were (a) av-
erage HR, SBP, and DBP for the resting baseline, stress task instruction, task
preparation, speech, mental arithmetic, and resting recovery periods; and (b)
salivary cortisol levels for the resting baseline, stress anticipation, poststress
and resting recovery time points. Greenhouse-Geisser adjusted degrees of
freedom were used when appropriate.
3
In both conditions, participants’ speeches were audiotaped. NON-EVAL
condition participants were told that this audiorecording would only be used
to ensure that the speeches given were on the correct topic but would not be
evaluated in any way. SOC-EVAL participants were told it would serve as an
additional source for evaluation. Audiorecording was used in order to encour-
age and confirm that participants spoke on the assigned topic.
4
In both conditions, sound vocalizations (but not actual spoken content)
were monitored to ensure that the participants spoke for the entire speech
period. The participant was warned that if he ceased speaking for a period of
more than 30 seconds, the experimenter would send a sound alert (via a
two-way radio) from the next room to signal the participant that he needed to
continue speaking. Sound monitoring was instituted to ensure that all partic-
ipants spoke for the full period in both stress conditions, so that there would
not be an imbalance of activity between participants or stress conditions. In
actuality, the alerting procedure was used for only 2 of the 81 participants, one
from each experimental condition (one subject alerted two times, one subject
alerted three times). The remainder of the 81 participants never ceased
speaking for longer than a 30-second period.
918
T. L. GRUENEWALD et al.
RESULTS
Baseline Differences
SOC-EVAL and NON-EVAL condition participants did
not differ on any baseline psychological or physiological
variables, except for anxiety. SOC-EVAL participants had
higher levels of baseline anxiety (M ⫽ 2.14, SEM ⫽ 0.13)
than NON-EVAL condition participants (M ⫽ 1.78, SEM ⫽
0.09), t(79) ⫽ 2.31, p ⫽ .02. As this difference in baseline
anxiety may have affected psychological and physiological
responses to the stressor tasks, baseline anxiety was used as a
covariate in repeated-measures ANCOVAs.5
There was a significant main effect of gender for SBP and
significant gender by time interactions for SBP and DBP in
repeated-measures ANOVAs, with males having higher BP
levels. However, gender did not interact with stress condition
to affect BP levels. There were no baseline or reactivity
gender differences for HR, cortisol, or any of the psycholog-
ical variables. Time of day analyses indicated that there was a
marginally significant trend for lower baseline cortisol levels
during the last timeslot of the day (5:30 –7:00 PM), as com-
pared with the preceding timeslots (2:30 – 4:00 and 4:00 –5:
30). However, time of day did not predict differential patterns
of cortisol responses across the session or interact with exper-
imental condition in analyses.
Manipulation Checks
There was not a significant difference in ratings of task
difficulty for participants in the social and nonsocial stress
conditions (SOC-EVAL M ⫽ 5.05, SEM ⫽ 0.21; NON-
EVAL M ⫽ 4.80, SEM ⫽ 0.22, t(79) ⫽ 0.82, p ⫽ .42).
SOC-EVAL condition participants rated the tasks as more
threatening (SOC-EVAL M ⫽ 2.80, SEM ⫽ 0.28, NON-
EVAL M ⫽ 2.03, SEM ⫽ 0.21, t(79) ⫽ 2.21, p ⫽ .03), and
marginally more challenging (SOC-EVAL M ⫽ 5.66, SEM ⫽
0.18; NON-EVAL M ⫽ 5.20, SEM ⫽ 0.18, t(79) ⫽ 1.80, p ⫽
.08). Mean scores for all five ratings of the panel of evaluators
for participants in the social condition were significantly
lower than the scale midpoint of 4 (all p’s ⬍ .05), indicating
that participants in the social condition did not like the pan-
elists or feel that the panelists would judge them or their
performance favorably.
Psychological Responses
Between- and within-subjects results from repeated-mea-
sures ANCOVAs for emotion and self-esteem variables are
presented in Table 1, and pre- and poststress means for each
variable in each condition are depicted in Figure 2. There was
a significant effect of time for each variable (SSGS shame
effect marginally significant), such that negative emotion
poststress scores were higher and poststress self-esteem scores
were lower than prestress scores. As hypothesized, only the
shame emotion subscales and the social self-esteem measure
5
Analyses for psychological and physiological variables not including
baseline anxiety as a covariate produced a similar pattern of results to those
reported here.
Psychosomatic Medicine 66:915–924 (2004)
SOCIAL-SELF THREAT
TABLE 1. Summary of Repeated-Measures ANCOVAs for Psychological Measures*

Anxiety Shame Shame Social Performance

Source
(ABS) F (ABS) F (SSGS) F SE F SE F

Between participants
Stress condition 7.74** 4.22* 1.15 3.81† 0.99
Anxiety — 24.24*** 10.05** 13.17*** 7.20**
Within participants
Time 94.98*** 13.17*** 2.72† 4.35* 14.51***
Time ⫻ anxiety — 2.39 0.51 0.03 0.21
Time ⫻ stress condition 0.16 8.01** 3.74† 7.44** 0.01

SOC-EVAL condition N ⫽ 41, NON-EVAL condition N ⫽ 40. Between- and within-subjects df: anxiety (1,77), ABS shame (1,73), SSGS shame (1,78), social
and performance self-esteem (1, 78).
*p ⬍ .05, **p ⬍ .01, ***p ⬍ .001, †p ⬍ .10.

showed significant stress condition by time effects (SSGS
shame effect marginally significant).6 Poststress shame scores
were higher (ABS shame Mdifference ⫽ 0.59, p ⫽ .01; SSGS
shame Mdifference ⫽ 0.28, p ⫽ .12), and social self-esteem
scores lower (mean difference ⫽ ⫺0.56, p ⫽ .01), for SOC-
EVAL stress condition participants as compared with NON-
EVAL stress condition participants.
Physiological Responses
Results of repeated-measures ANCOVAs for HR, SBP, and
DBP are displayed in Table 2, and mean values for each
measurement time are depicted in Figure 3. There was a
significant effect of time for all three cardiovascular measures,
with significantly higher than baseline values at the task
instruction, task preparation, and speech and math task mea-
surement points (all p’s ⬍ .01). Recovery BP values did not
significantly differ from baseline values, while HR recovery
values were lower than baseline values (p ⬍ .001). There were
significant stress condition by time interactions for all three
variables. Compared with NON-EVAL stress condition par-
ticipants, subjects in the SOC-EVAL stress condition had
marginally higher HR and SBP values at the speech measure-
ment time point but marginally lower DBP values at baseline
and marginally lower SBP and DBP values at the task instruc-
tion time points (all p’s between .05 and .10).
Results of repeated-measures ANCOVAs for salivary cor-
tisol are displayed in Table 2, and baseline, stress anticipation,
poststress, and resting values for the SOC-EVAL and NON-
EVAL stress condition participants are presented in Figure 4.
There was a significant between-subjects effect of stress con-
dition, which was qualified by a significant stress condition by
time interaction. Baseline and stress anticipation cortisol lev-
els did not differ between SOC-EVAL and NON-EVAL stress
condition participants. As hypothesized, salivary cortisol lev-
els were significantly higher at the poststress (Mdifference ⫽
0.12, p ⫽ .002) and recovery time points (Mdifference ⫽ 0.09,
6
Results for other subscales of the Affect Balance Scale are not discussed
here. However, there were significant effects of time for depressed emotion,
affection, and contentment (poststress negative emotion higher and positive
emotions lower), but not for anger, joy, or vigor. There were no significant
stress condition by time interactions for any of these other emotion subscales.
p ⫽ .002), in the SOC-EVAL as compared with the NON-
EVAL stress condition.
To assess whether participants in the SOC-EVAL condition
who showed high or low changes in emotion and self-esteem
variables exhibited differential increases in pre- to poststress
cortisol levels, median splits were performed on change scores
for each emotion and self-esteem variable, and the mean
change in cortisol for low and high change groups was exam-
ined (residualized changed scores were computed by regress-
ing baseline value onto poststress value). Standardized pre- to
poststress cortisol change for SOC-EVAL condition partici-
pants showing above vs. below the median levels of self-
esteem decreases and emotion increases are displayed in Fig-
ure 5. Participants showing above the median levels of social
self-esteem decrease and shame increase had higher levels of
pre- to post-stress cortisol increase (social self-esteem, t(38) ⫽
⫺2.02, p ⫽ .05; shame (ABS), t(36) ⫽ 2.05, p ⫽ .05; shame
(SSGS), nonsignificant, t(38) ⫽ 1.55, p ⫽ .13). There were no
significant differences in mean levels of cortisol change in
high and low change groups for performance self-esteem,
t(38) ⫽ ⫺0.11, p ⫽ .91, and anxiety, t(38) ⫽ ⫺0.63, p ⫽ .53.
DISCUSSION
The present investigation tested the theoretical premise of
our social-self preservation model that situations that threaten
the social self elicit increases in shame, decrements in social
self-esteem, and increases in cortisol by comparing psychobi-
ological responses to performance stressor tasks under the
presence or absence of direct social-evaluative threat. Perfor-
mance of the tasks in both stress conditions appeared to be
“stressful” as indexed by ratings of the tasks as difficult and
increases in anxiety and decreases in performance self-esteem.
However, those in the social-evaluative condition showed
higher poststress shame and lower social state self-esteem
than those who performed the tasks without direct evaluative
threat. Cortisol levels did not increase in response to perfor-
mance of stress tasks in the nonevaluative condition, but as
hypothesized a large poststress increase occurred in response
to performance of these activities under social-evaluative
threat. Although we found significant time by stress condition
effects for the cardiovascular system indicators, changes in

Psychosomatic Medicine 66:915–924 (2004) 919


Figure 2. Mean (⫾ SE) pre- and post-stressor scores on emotion and self-esteem measures in SOC-EVAL and NON-EVAL stressor conditions.
these parameters did not seem to be as sensitive to the social
characteristics of stressor experience. HR and BP showed
large increases in both stressor conditions, and HR and SBP
were only marginally greater in the social evaluation condition
at the speech measurement point.
Implications for Stress Research
The pattern of responses in the emotion, self-esteem and
task rating measures indicates that both stressor condition
920
T. L. GRUENEWALD et al.
groups found the tasks to be “stressful” in accordance with
standards typically used in acute stress research (eg, per-
ceived task difficulty, increased anxiety). However, perfor-
mance of these demanding tasks in the presence of an
evaluative social audience led to greater responses in the
specific set of psychological states that we believe are
indicative of a threat to the social self: increases in shame
emotions and decreases in the social component of self-
esteem. Greater cortisol increases were observed in partic-
Psychosomatic Medicine 66:915–924 (2004)
SOCIAL-SELF THREAT
TABLE 2. Summary of Repeated-Measures ANCOVAs for HR, SBP, DBP, and Cortisol*
HR
Source
F F
Between participants
Stress condition 1.36
Anxiety 0.03
Within participants
Time 14.44***
Time ⫻ anxiety 1.40
Time ⫻ stress condition 3.41*
SOC-EVAL condition N ⫽ 41, NON-EVAL condition N ⫽ 40. HR, SBP, and DBP between-subjects df: (1,73); cortisol between subjects df (1,75); F-values
with Greenhouse-Geisser adjusted within-subjects df reported above: HR df (2.66,194.04), SBP df (3.98,290.23), DBP df (3.62,264.43), cortisol df (1.33,99.79).
*p ⬍ .05, **p ⬍ .01, ***p ⬍ .001, †p ⬍ .10.
ipants who experienced greater increases in shame and
greater decreases in social self-esteem, but cortisol in-
creases did not significantly differ as a function of the
magnitude of increases in anxiety or decreases in perfor-
mance self-esteem. The lack of an association between
increases in cortisol and anxiety conflicts with other inves-
tigations that have found associations between stress-in-
duced increases in these parameters (41,42); however, these
associations are not always found in acute stress research
(43,44), nor are such associations always examined. The
finding of greater concordance between large stress-in-
duced increases in cortisol and shame complements find-
ings of a previous investigation which documented greater
cortisol increases in response to task failure in children who
experienced more postfailure shame, embarrassment, and
negative self-evaluation (45). Thus, threats to the social self
and the emotions and cognitions that derive from such
threats may be important elicitors of this critical physio-
logical system. More research will be needed to thoroughly
elucidate whether other cognitive and emotional processes
not examined in the present investigation may underlie
greater cortisol increases witnessed under social-evaluative
threat, including psychological factors thought to be linked
to HPA activity, such as perceived control, novelty, unpre-
dictability, and ego involvement, and to more clearly de-
lineate links between cortisol reactivity and changes in
shame and social self-esteem.
In this investigation, we employed a version of the TSST
developed by Kirschbaum and his colleagues (35,40), a par-
adigm that has been employed in many acute stress investi-
gations and is known to reliably elicit cortisol increases. Our
results indicate that threat to the social self inherent in this
paradigm may be an important “active ingredient” in the
elicitation of a cortisol increase. This investigation also ex-
perimentally confirms the results of Dickerson and Kemeny’s
meta-analysis (36), which demonstrated that social-evaluative
threat within stressor paradigms was associated with cortisol
reactivity. A long history of theoretical and empirical research
has documented the types of threats likely to activate the HPA
axis, such as threats to physical integrity, fear-provoking
events, and events characterized by unpredictability, novelty,
Psychosomatic Medicine 66:915–924 (2004)
SBP DBP Cortisol
F F
0.00 0.49 5.44*
0.04 1.12 3.04†
15.36*** 15.78*** 0.46
1.72 0.92 2.74†
6.91*** 3.17* 12.57***
and loss of control (24 –27). Our goal in formulating the Social
Self Preservation Theory is to encourage attention to threat to
the social self as an additional important elicitor of this sys-
tem, as maintaining the integrity of the social self may also be
important for survival and well-being. Chronic exposure to
social self threats, such as that which occurs in those with a
stigmatizing condition or those consistently subjected to or
particularly sensitive to rejection, might provide a consistent
trigger to this system, resulting in increased risk for diseases
linked to this system (see 22,46).
Implications for Emotion Research
Negative emotional and cognitive responses are hypothe-
sized to be important pathways between stressor experience
and adverse physiological and health states. However, the
majority of research in this area has focused on the physio-
logical correlates and health consequences of negative mood
(a composite of a number of negative moods) or depressed
mood. Less attention has been given to specific emotions in
this research (with the exception of anger and cardiovascular
activity). The present findings support the utility of examining
specific emotional and cognitive responses to different char-
acteristics of stressful situations; such efforts may help to
clarify our understanding of associations between stressor
experience and health outcomes (47).
Many theories of the etiology of shame emotions and
social self-esteem hypothesize that these states are partic-
ularly likely to occur in social situations in which the self
is devalued (6,8,14); however, little empirical work in
which aspects of the social environment are manipulated to
test this hypothesis exists. The present investigation pro-
vides experimental support for the differential experience
of shame and social self-esteem as a function of the social
context of stressful situations. Future research further de-
lineating changes in specific feelings of low social worth
(eg, shame, embarrassment, humiliation) and other catego-
ries of negative emotion, as well as concomitant changes in
different physiological parameters, to different types of
social and nonsocial stressors is needed.
921

Figure 3. Mean (⫾ SE) HR, SBP, and DBP values in SOC-EVAL and NON-EVAL stressor conditions across the session.
Limitations and Future Directions
We included HR and blood pressure measures to assess
increases in general levels of arousal that we thought would
result from performance of tasks equally in both stress con-
ditions. The finding of marginally higher HR and SBP re-
sponses in evaluative condition participants during the speech
portion of the stress activities does indicate that these param-
eters are sensitive to social-evaluative characteristics of stres-
sor performance, a finding consistent with an investigation
that found greater SBP increases to a public speaking stressor
in a condition in which social evaluation was emphasized
compared with one in which it was not (48). Perhaps if we had
922
T. L. GRUENEWALD et al.
employed additional and more comprehensive measures of
cardiovascular activity, such as impedance cardiography and
electrocardiographic techniques, we might have been better
able to detect differences in autonomic response between our
two stress conditions. Research with such techniques has
indicated differential effects of social evaluation during per-
formance of learned and unlearned tasks on cardiac and vas-
cular components of autonomic reactivity (49) and differential
effects of the introduction of evaluative observation on cardiac
and vascular response adaptation to repeated performance
stressor exposure (50). Thus, future investigations that include
such techniques will be useful in clarifying our understanding
Psychosomatic Medicine 66:915–924 (2004)
SOCIAL-SELF THREAT

Figure 4. Mean (⫾ SE) salivary cortisol values in SOC-EVAL and NON-EVAL stressor conditions across the session.

Figure 5. Mean (⫾ SE) pre- to poststressor salivary cortisol increases in “above median change” social self-esteem and shame groups (ie, those participants
that showed changes in social self-esteem and shame from pre- to poststressor that were above the median) and “below median change” groups (ie, those
participants that showed changes in these psychological responses that were below the median) in the SOC-EVAL stressor condition.

of the activity of these systems under conditions of threat to
the social self. Future research will also be needed to clarify
the factors that account for the larger cardiovascular responses
witnessed during the speech as compared with the math task in
both conditions, which is an outcome that has been demon-
strated in a previous investigation (41). Probable candidates
include differences in physical activity (eg, participants stood
during speech task and sat during math task) and psychosocial
factors (eg, perhaps speech task more ego involving). The
more pronounced differences in HR and SBP at the speech
time point between the two stress conditions also suggests that
speaking tasks may elicit greater concerns regarding others’
evaluations of the social self.
We chose to test our theory in this first investigation by
manipulating the presence or absence of social-evaluative
threat within a commonly used performance stress para-
digm, so that our findings could be compared with previous
acute stress investigations and so that the role of the social
context of stressor experience in the elicitation of specific
emotional, cognitive, and physiological responses would be
highlighted. It could be argued, however, that manipulating
the presence or absence of social-evaluative threat while
holding all other activities constant may have produced two
experimental conditions that varied in intensity (ie, the
social-evaluative condition simply “adds” social-evaluation
to a performance stressor, yielding “more stress”) rather
than in nature (ie, the presence or absence of social-eval-
uative threat). Although we cannot rule out this alternative
explanation for our results, we do feel it is important to
highlight that only certain psychological and physiological
responses varied significantly across the two stress condi-
tions. If our social-evaluative condition was simply more
intense, it might be expected that most of the psychological
and physiological stress response measures (eg, anxiety,
performance self-esteem) would have shown greater
changes in this condition. It will be important to address
this issue of difference in intensity versus difference in the
nature of a stressor in future research.
Although we believe that performance of speech and math
tasks under social-evaluative threat is a provocative situation,
capable of eliciting concerns about the status of one’s social self,
it will be important to examine psychological and physiological
responses to other forms of threat to the social self, including
those outside the laboratory and those that may be common
outside the college environment. Given the potential ubiquity of
acute and chronic social self threats in our environment, a greater
understanding of the psychobiological consequences of such
threats is an important aim of future research.

Psychosomatic Medicine 66:915–924 (2004) 923


We thank Sally Dickerson, Shelley Taylor, Teresa Seeman, and Neil
Malamuth for their helpful comments on the rationale and design of
this research. We are indebted to Julie Bower, Roberta Mancuso, and
Shelley Taylor’s 421 Research Group for their comments and sug-
gestions on previous versions of this manuscript. We thank Dennis
Montoya for his assistance with the assays for salivary cortisol.
REFERENCES
1. Dickerson SS, Gruenewald TL, Kemeny ME. When the social self is
threatened: shame, physiology, and health. J Pers 2004;72:1191–216.
2. Baumeister RF, Leary MR. The need to belong: desire for interpersonal attach-
ments as a fundamental human motivation. Psych Bull 1995;117:497–529.
3. James W. The principles of psychology. Chicago: Encyclopaedia
Britannica; 1952.
4. Bowlby J. A secure base: parent-child attachment and healthy human
development. New York: Basic Books; 1988.
5. Barkow JH. Darwin, sex, and status. Toronto: University of Toronto
Press; 1989.
6. Gilbert P. Shame, status, and social roles: psychobiology and evolution.
In: Gilbert P, Andrews B, editors. Shame: interpersonal behavior, psy-
chopathology, and culture. New York: Oxford University Press; 1998.
7. Darwin C. The expression of the emotions in man and animals. Chicago:
University of Chicago Press; 1872/1965.
8. Gilbert P. The evolution of social attractiveness and its role in shame,
humiliation, guilt and therapy. Br J Med Psychol 1997;70(pt 2):113– 47.
9. Lewis HB. Shame and guilt in neurosis. Psychoanal Rev 1971;58:419 –38.
10. Lewis M. Self-conscious emotions: embarrassment, pride, shame, and
guilt. In: Lewis M, Haviland JM, editors. Handbook of emotions. New
York: Guilford Press; 1993.
11. Tangney JP, Fischer KW. Self-conscious emotions: the psychology of
shame, guilt, embarrassment, and pride. New York: Guilford Press; 1995.
12. Cooley CH. Human nature and the social order. New Brunswick: Trans-
action Books; 1902/1983.
13. Fessler DMT. Toward an understanding of the universality of second-
order emotions. In: Hinton AL, editor. Biocultural approaches to the
emotions. New York: Cambridge University Press; 1999.
14. Scheff TJ. Shame and conformity: the deference-emotion system. Am
Sociol Rev 1988;53:395– 406.
15. Tangney JP, Miller RS, Flicker L, Barlow DH. Are shame, guilt, and
embarrassment distinct emotions? J Pers Soc Psychol 1996;70:1256 – 69.
16. Wicker FW, Payne GC, Morgan RD. Participants’ descriptions of guilt
and shame. Mot Emotion 1983;7:25–39.
17. Keltner D, Buswell BN. Evidence for the distinctness of embarrassment,
shame, and guilt: a study of recalled antecedents and facial expressions of
emotion. Cogn Emotion 1996;10:155–71.
18. Tangney JP. Situational determinants of shame and guilt in young adult-
hood. Pers Soc Psychol Bull 1992;18:199 –206.
19. Leary MR, Tambor ES, Terdal SK, Downs DL. Self-esteem as an
interpersonal monitor: the sociometer hypothesis. J Pers Soc Psychol
1995;68:518 –30.
20. Leary MR, Haupt AL, Strausser KS, Chokel JT. Calibrating the
sociometer: the relationship between interpersonal appraisals and state
self-esteem. J Pers Soc Psychol 1998;74:1290 –9.
21. Selye H. The physiology and pathology of exposure to stress: a treatise
based on the concepts of the general-adaptation-syndrome and the dis-
eases of adaptation. Montreal: Acta; 1950.
22. McEwen BS, Biron CA, Brunson KW, Bulloch K, Chambers WH,
Dhabhar FS. The role of adrenocorticoids as modulators of immune
function in health and disease: neural, endocrine and immune interac-
tions. Brain Res Rev 1997;23:79 –133.
23. Sapolsky RM. Neuroendocrinology of the stress-response. In: Becker JB,
Breedlove SM, Crews D, editors. Behavioral endocrinology. Cambridge,
MA: MIT Press; 1992.
24. Lovallo WR, Thomas TL. Stress hormones in psychophysiological
research: emotional, behavioral, and cognitive implications. In: Cacioppo
924
T. L. GRUENEWALD et al.
JT, Tassinary LG, Bernston GG, editors. Handbook of psychophysiology.
2nd ed. Cambridge University Press; 2000.
25. Kirschbaum C, Hellhammer DH. Salivary cortisol in psychobiological
research: an overview. Neuropsychobiology 1989;22:150 – 69.
26. Biondi M, Picardi A. Psychological stress and neuroendocrine function in
humans: the last two decades of research. Psychother Psychosom 1999;
68:144 –50.
27. Mason JW. A review of psychoendocrine research on the pituitary-
adrenal cortical system. Psychosom Med 1968;30:576 – 607.
28. Stroud LR, Salovey P, Epel ES. Sex differences in stress responses: social
rejection versus achievement stress. Biol Psychiatry 2002;52:318 –27.
29. Gunnar MR, Donzella B. Social regulation of the cortisol levels in early
human development. Psychoneuroendocrinology 2002;27:199 –220.
30. Seeman TE, McEwen BS. Impact of social environment characteristics
on neuroendocrine regulation. Psychosom Med 1996;58:459 –71.
31. Avitsur R, Stark JL, Sheridan JF. Social stress induces glucocorticoid
resistance in subordinate animals. Horm Behav 2001;39:247–57.
32. Cohen S, Line S, Manuck SB, Rabin BS, Heise ER, Kaplan JR. Chronic
social stress, social status, and susceptibility to upper respiratory infec-
tions in nonhuman primates. Psych Med 1997;59:213–21.
33. Sapolsky RM. Endocrinology alfresco: psychoneuroendocrine studies of
wild baboons. Recent Prog Horm Res 1993;48:437– 68.
34. Shively CA, Laber-Laird K, Anton RF. Behavior and physiology of
social stress and depression in female cynomolgus monkeys. Biol Psy-
chiatry 1997;758:871– 82.
35. Kirschbaum C, Pirke KM, Hellhammer DH. The “Trier Social Stress
Test”: a tool for investigating psychobiological stress responses in a
laboratory setting. Neuropsychobiology 1993;28:76 – 81.
36. Dickerson SS, Kemeny ME. Acute stressors and cortisol responses: a
theoretical integration and synthesis of laboratory research. Psychol Bull
2004;130:355–91.
37. Derogatis LR. Affects Balance Scale. Riderwood, MD: Clinical Psycho-
metric Research; 1975.
38. Marschall DE, Sanftner JL, Tangney JP. The State Shame and Guilt
Scale. Fairfax, VA: George Mason University; 1994.
39. Heatherton TF, Polivy J. Development and validation of a scale for
measuring state self-esteem. J Pers Soc Psychol 1991;60:895–910.
40. Pruessner JC, Hellhammer DH, Kirschbaum C. Low self-esteem, induced failure
and the adrenocortical stress response. Pers Individ Dif 1999;27:477–89.
41. Al’Absi M, Bongard S, Buchanan T, Pincomb GA, Licinio J, Lovallo
WR. Cardiovascular and neuroendocrine adjustment to public speaking
and mental arithmetic stressors. Psychophysiology 1997;34:266 –75.
42. Alpers GW, Abelson JL, Wilhelm FH, Roth WT. Salivary cortisol re-
sponse during exposure treatment in driving phobics. Psychosom Med
2003;65:679 – 87.
43. Schedlowski M, Jacobs R, Stratmann G, Richter S, Hadicke A, Tewes U.
Changes of natural killer cells during acute psychological stress. J Clin
Immunol 1993;13:119 –26.
44. Schedlowski M, Wiechert D, Wagner TOF, Tewes U. Acute psycholog-
ical stress increases plasma levels of cortisol, prolactin, and TSH. Life Sci
1992;50:1201–5.
45. Lewis M, Ramsay D. Cortisol response to embarrassment and shame.
Child Dev 2002;73:1034 – 45.
46. McEwen BS, Seeman TE. Protective and damaging effects of mediators
of stress: elaborating and testing the concepts of allostatis and allostatic
load. Ann N Y Acad Sci 1999;896:30 – 47.
47. Kemeny ME. The psychobiology of stress. Psych Sci 2003;12:124 –9.
48. Smith TW, Nealey JB, Kircher JC, Limon JP. Social determinants of
cardiovascular reactivity: effects of incentive to exert influence and
evaluative threat. Psychophysiology 1997;34:65–73.
49. Blascovich J, Mendes WB, Hunter SB, Salomon K. Social facilitation as
challenge and threat. J Pers Soc Psychol 1999;77:68 –77.
50. Kelsey RM, Blascovich J, Leitten CL, Schneider TR, Tomaka J, Wiens S.
Cardiovascular reactivity and adaptation to recurrent psychological
stress: the moderating effects of evaluative observation. Psychophysiol-
ogy 2000;37:748 –56.
Psychosomatic Medicine 66:915–924 (2004)