SOCIAL BEHAVIOR AND PERSONALITY, 2006, 34(3), 221-238

© Society for Personality Research (Inc.)

INVESTIGATING THE COMPASS OF SHAME: THE

DEVELOPMENT OF THE COMPASS OF SHAME SCALE

JEFF ELISON
University of Denver, CO, USA
RANDY LENNON AND STEVEN PULOS
University of Northern Colorado, CO, USA

The way in which one copes with, or defends against, shame has important implications.
The Compass of Shame Scale (CoSS) was developed to assess use of the four shame-
coping styles described by Nathanson (1992): Attack Self, Withdrawal, Attack Other, and
Avoidance. Reliability and criterion validity were explored (N = 322). Subscale reliabilities
ranged from .74 to .91. Confirmatory factor analysis supported a model with four primary
factors. A differentiated pattern of correlations was obtained between CoSS scales and
general internalized shame, self-esteem, anger, coping, and psychological symptoms. Results
provided empirical support for Nathanson’s Compass of Shame model and the validity of the
CoSS.

Keywords: shame, coping, psychopathology, assessment, validity

An important aspect of the experience of shame is the way in which one copes
with, or defends against, it. Shame is a painful, self-focused affect (Parker, 1998),
which has been linked to many problematic psychological states including:

Jeff Elison, formerly of the University of Denver, CO, USA, now of the Department of Psychology,
Southern Utah University, UT, USA; Randy Lennon and Steven Pulos, University of Northern
Colorado, CO, USA
Portions of this study were presented at the Society for Personality Assessment 2001 Midwinter
Meeting, March 2001, Philadelphia and at the Rocky Mountain Psychological Association 2001
Meeting, April 2001, Reno.
The authors would like to thank Susan Harter, Donald Nathanson, and Justin Campbell for their
helpful comments on earlier drafts of this paper.
Appreciation is due to reviewers including: Sally Dickerson, PhD, Department of Psychology &
Social Behavior, University of California, 3340 Social Ecology II, Irvine, CA 92697-7085, USA.
Email: [email protected]
Please address correspondence and reprint requests to: Jeff Elison, Department of Psychology,
Southern Utah University, 351 W. Center St., Cedar City, UT 84720, USA. Phone: +01 435-586-7889;
Fax: +01 435 865 8289; Email: [email protected]

221
222 COMPASS OF SHAME SCALE

aggression, depression, somaticization, anxiety, obsessive-compulsive disorder,

interpersonal sensitivity, personality disorders, substance abuse, posttraumatic
stress disorder, and low self-esteem (e.g., Harder, 1995; Lewis, 1971, 1987;
Nathanson, 1992, 1994; Tangney & Dearing, 2002). Recent reviews of the shame
and guilt literature encourage investigations of the context under which these
emotions are adaptive versus maladaptive (Bybee & Quiles, 1998; Ferguson &
Stegge, 1998; Tangney & Dearing, 2002). While many assessments of the trait of
shame proneness have been developed (e.g., Cook, 2001; Harder & Zalma, 1990;
Tangney, Wagner, & Gramzow, 1989), the Compass of Shame Scale, described
here, is the first instrument specifically designed to assess shame-coping styles.
Nathanson (1992) proposed a model for shame-management scripts or coping
styles, the Compass of Shame. This model describes four families of scripts,
represented by the poles of the compass and labeled Attack Self, Withdrawal,
Attack Other, and Avoidance. Each set of shame-focused scripts is associated
with different motivations, affects, cognitions, behaviors. Clinicians adopting
Tomkins’s affect theory (1962 - 1991) have used Nathanson’s model as a guide
in psychotherapeutic practice (e.g., Nathanson, 1994, 2003). The present study
investigates the reliability and validity of an instrument, the Compass of Shame
Scale (CoSS), to measure Nathanson’s (1992) Compass of Shame model.

THE COMPASS OF SHAME

The experience of shame is central to Nathanson’s (1992) Compass of Shame
model; coping scripts are triggered in reaction to a shaming event. Nathanson,
following Tomkins, uses a very broad definition of shame, that is, shame is the
negative affect felt in response to any impediment to the ongoing experience
of interest or joy (Tomkins, 1963). Nathanson presents his Compass of Shame
model within the framework of Tomkins’s script theory (1991). Scripts, much
like schemas, are recursively defined and nested; they are “sets of ordering
rules for the interpretation, evaluation, prediction, production, or control of
scenes” (Tomkins, 1991, p. 84). A constructive script for shame management is
to attend to its source and evaluate whether or not one cares to address it; few
people consistently achieve this ideal. The four poles of the Compass of Shame
characterize the many scripts by which shame is reduced, ignored, or magnified,
without addressing its source (Nathanson, 1992).
At the Withdrawal pole, the person acknowledges the experience as negative,
accepts shame’s message as valid, and tries to withdraw or hide from the situation.
For example, a shamed student might decide not to participate in class discussions
or in the extreme case may even drop out of school. The phenomenological
experience is negative; emotions include shame, sadness, fear, and anxiety.
Cognitions include awareness of one’s discomfort with others, and possibly
awareness of shameful actions, faults, or characteristics. Nevertheless, negative
 COMPASS OF SHAME SCALE 223
feelings and cognitions may not be identified explicitly as shame. The motivation
is to limit shameful exposure via the action tendency of withdrawing.
At the Attack Self pole, the person acknowledges the experience as negative,
accepts shame’s message as valid, and turns anger inward. For example, a
shamed student might feel self-directed rage for being “stupid.” The phe-
nomenological experience is negative; emotions include self-directed anger,
contempt, or disgust, which magnify the impact of shame. Cognitions include
awareness of one’s shameful actions, faults, or characteristics. As in Withdrawal,
negative feelings and cognitions may be acknowledged, but may not be identified
explicitly as shame. The motivation is to take control of shame with the ultimate
goal being to win acceptance by others. The action tendency is to criticize the
self, prevent reoccurrence of the shameful situation through change, conform,
show deference to others, or engage in self-deprecating remarks.
Withdrawal and Attack Self share two important aspects, recognition of the
negative experience and conscious acceptance (internalization) of shame’s
message; the self is found lacking. However, it is important to note that while
individuals may acknowledge the negative experience (e.g., “I feel bad”) and
accept shame’s message (e.g., “I’m worthless,” “I hate myself”), they may
not explicitly identify the experience or feeling as shame per se. An important
difference between the two poles can be seen in their motivations; individuals
using Attack Self scripts often endure shame in order to maintain relationships
with others; while those using Withdrawal scripts pull away from others in order
to reduce their discomfort and shame experiences.
At the Avoidance pole, the person typically does not acknowledge the negative
experience of self, typically does not accept shame’s message as valid (denial),
and attempts are made to distract the self and others from the painful feeling. For
example, a shamed student might joke about a failing grade or disavow interest in
the class. The phenomenological experience becomes neutral or positive; shame
may be disavowed, or overridden with joy or excitement via distractions (e.g.,
sex). Cognitions include little awareness of shame or one’s shameful actions,
faults, or characteristics. The motivation is to minimize the conscious experience
of shame or show oneself as being above shame. Of all the poles, Avoidance
scripts are most likely to operate outside of consciousness.
Finally, at the Attack Other pole, the person may – or may not – acknowledge
the negative experience of self, typically does not accept shame’s message, and
attempts are made to make someone else feel worse. For example, a shamed
student might turn the tables by teasing someone else or externalize the shame
by blaming the teacher. The phenomenological experience is negative; anger is
directed outward, perhaps toward the source of the shaming event. The cognitive
experience is an awareness of someone else’s actions or faults and may, or may
not, involve awareness of shame. The motivation is to bolster one’s own self-
224 COMPASS OF SHAME SCALE

image and externalize the shame. The action tendency is to verbally or physically
attack someone or something else in order to make someone else feel inferior.
Attack Other and Avoidance share the minimization and limited awareness of
shame.
Thus, the poles of the compass can be ordered according to the degree to which
they involve consciousness and internalization of shame: Withdrawal and Attack
Self are equal, both being greater than Attack Other, which is in turn greater than
Avoidance. This observation underlies many hypotheses involving differential
relationships between these shame-coping styles and other variables of interest,
including the frequent lack of discrimination between Withdrawal and Attack
Self.
The poles of the compass model may be viewed from either a state or trait
perspective, similar to the way in which anger may be viewed as either a state or
a trait. As a state, a given shame-coping script may be active briefly, for seconds
at a time, or active over a longer period. Multiple scripts may be employed in
coping with a single shaming event, consistent with the literature on coping
(Horowitz, 1979; Lazarus, 1999; Nathanson, 1992). As a trait, which poles
of the compass are employed, frequencies of use, and possibly order of use,
are characteristic of an individual over time (Elison, Pulos, & Lennon, 2006).
Because multiple scripts may be employed in response to a single situation,
and because preference of use may be stable, the poles of the compass are not
necessarily independent. For example, acceptance of shame is shared by Attack
Self and Withdrawal; minimization is shared by Attack Other and Avoidance;
anger is shared by Attack Self and Attack Other. Thus, an individual may tend
to employ scripts from specific pairs of poles together; their application may be
successive or simultaneous.

THE DEVELOPMENT OF THE COMPASS OF SHAME SCALE

The Compass of Shame Scale (CoSS) was developed to assess an individual’s
use of the four styles of shame reactions described by Nathanson’s (1992)
Compass of Shame model. Development began with a blueprint, or table of
specifications, for test item development. Material for the blueprint came from
Shame and Pride (Nathanson, 1992).
The format of the CoSS is a series of items the stems of which describe
potentially shame-inducing situations or variations of shame emotions (e.g.,
guilt, embarrassment, rejection) followed by four responses, representing
reactions characteristic of each of the poles. Nathanson’s eight categories of
shame-inducing situations described in Shame and Pride (1992, p. 317), were
used for the item stems. Shame and Pride contains a chapter describing each
pole of the compass; these were consulted for responses. Due to concern over
the ability to assess shame responses via self-report, the responses do not require
 COMPASS OF SHAME SCALE 225
acknowledgement of shame per se. For example, rather than inquiring about
shame directly, responses describe related feelings and behaviors such as getting
angry at others, being self-critical, or making jokes. Individuals are instructed to
rate every item using a forced-choice frequency rating, ranging from 0 (never)
to 4 (almost always).
The structure of four responses for each stem, all of which are rated
independently, was chosen for four reasons. First, the use of multiple responses
reflects the literature on coping, in that multiple coping mechanisms (scripts or
schemas) may be employed in reaction to a single situation (Horowitz, 1979;
Lazarus, 1999; Nathanson, 1992). Second, participants rate every response,
rather than order them, because an ipsative scale was not desired. In other words,
participants can – and do – rate all four responses with the same frequency if
they desire. Third, having four responses for every stem allows ratings to be
summed for each stem in order to compare ratings among the stem scenarios.
This type of comparison by scenario is of interest between groups as well
as within individuals. The latter is of particular interest in therapy where the
CoSS can be used to illuminate which scenarios are particularly painful for an
individual (e.g., Cook, 2001). Fourth, grouping responses by stem rather than
repeating each stem four times reduces the length and repetitiveness of the CoSS.
The entire instrument (Version 3) appears in the Appendix with items annotated
to indicate their respective scales.
For the pilot (Version 1), 72 items were written from the blueprint. Four
members of the Tomkins Institute reviewed the items for content validity; eight
items were rewritten. Thirty-four undergraduate and graduate students completed
the CoSS and provided written feedback on the format, stem situations, and
responses. Data from the pilot demonstrated alpha reliabilities across the scales
of .82 to .88. Feedback indicated the 72-item version was long and repetitive for
many participants. In addition, they felt a number of responses were unlikely or
confusing. These items tended to have highly skewed distributions or low item-
total correlations or both. The items to remove were identified and the CoSS was
trimmed to 48 items for Version 2. Version 2 had acceptable reliability and factor
structure, but three of the items did not load as predicted (Elison et al., 2006).
These items were rewritten and the CoSS Version 3 was developed.
The current study examines the reliability and validity of the CoSS. Specific
issues include: 1) the reliability of the scales, 2) whether the internal structure
of the CoSS is consistent with the theoretical model, and 3) convergent and
discriminant validity of the scales as they relate to theoretically predicted
criterion variables. As with any validity study, positive findings also serve as
empirical support for the theoretical model on which the instrument is based.
226 COMPASS OF SHAME SCALE

METHOD

PARTICIPANTS AND PROCEDURE

The revised pool of 48 items (CoSS Version 3) and criterion measures were
administered to a sample of 322 students (83% undergraduate, 17% graduate; M
= 24.3 years old, SD = 8.4; 68% women, 32% men; 92% European-American).
Although all participants completed the CoSS, subgroups completed the criterion
measures (Internalized Shame Scale, n = 50; Attack Other Scale, n = 47; Ways
of Coping Questionnaire, n = 47; Symptom Check List, n = 150). Twenty-seven
participants were retested three weeks later to assess stability of the scores over
time.

CRITERION MEASURES
The Internalized Shame Scale (ISS; Cook, 2001) was used to explore convergent
and discriminant validity in relation to shame and self-esteem. The ISS contains
a 24-item shame scale and a 6-item self-esteem scale, referred to here as the
ISS-Shame and the ISS-SE, respectively. The coefficient alpha for the ISS shame
scale was .95 and alpha for the ISS self-esteem scale was .90.
The SCL-90-R (Symptom Check List; Derogatis, 1983) is a 90-item self-report
symptom inventory designed to assess patterns of psychological symptoms.
The SCL-90 contains nine symptom scales, and three summary scales. Three
symptom (Depression, Interpersonal Sensitivity, and Hostility-Anger) and
two total (General Severity Index (GSI) and Positive Symptom Distress Index
(PSDI)) scales were employed in the current study. The GSI is an index of overall
psychological distress while the PSDI is a measure of symptom intensity in that
it is corrected for the number of symptoms. The coefficient alphas for the SCL-
90-R scales used in the current study were: GSI, .98; PSDI, .98; Depression, .91;
Interpersonal Sensitivity, .83; Hostility-Anger, .79.
The Attack Other Scale (AOS; Cook, 1996) is a measure of other-directed anger
as a trait. Its nine items present statements about an individual’s anger, especially
in response to interpersonal situations. In the current study the coefficient alpha
for the AOS was .83.
Ways of Coping Questionnaire The Ways of Coping Questionnaire (WCQ;
Folkman & Lazarus, 1988) was used to explore convergent and discriminant
validity in relation to emotion-focused and problem-focused coping styles. The
WCQ is a 66-item instrument with subscales designed to assess an individual’s
use of eight different coping processes. These coping processes can be organized
into two higher order categories, problem-focused coping and emotion-focused
coping (Folkman, Lazarus, Dunkel-Schetter, DeLongis, & Gruen, 1986).
The CoSS is intended to measure the use of affect management scripts
(Tomkins, 1991); that is to say, the CoSS is a measure of emotion-focused
 COMPASS OF SHAME SCALE 227
coping, not problem-focused coping. These emotion-focused scales were used
to assess the convergent validity of the CoSS, while the problem-focused scales
were used to assess the divergent validity of the CoSS, that is, some CoSS scales
should correlate with other scales measuring emotion-focused coping and should
not generally correlate with scales measuring problem-focused coping.
Newer research suggests that Confrontive coping may be more emotion
focused and Positive Reappraisal may be more problem focused (Wei, Lennon,
& Pulos, 2005). Accordingly, Confrontive Coping, which includes hostility and
anger, should correlate with Attack Other, and Positive Reappraisal may not
correlate with any of the CoSS scales.

RESULTS

CONFIRMATORY ANALYSIS OF COSS STRUCTURE

A confirmatory factor analysis was conducted to examine whether the data
fit the a priori model of the CoSS. Specifically, the model predicted that each
item would be related only to the scale it was hypothesized to represent. Such a
model is a simultaneous test of the unidimensionality of the four CoSS scales.
The analysis was conducted with LISREL 8.54, and the Goodness of Fit was
evaluated with the two most widely accepted and employed indices of fit – the
Comparative Fit Index (CFI) and the Root Mean Square Error of Approximation
(RMSEA) (McDonald & Ho, 2002).
The CoSS model demonstrated acceptable fit to the data as the CFI was .94
(CFI greater than .90 is considered acceptable). The RMSEA was .055 with a
90% confidence interval between 0.052 and 0.059. An RMSEA with the upper
limit of the confidence interval below .10 is considered an acceptable fit.
Moderate correlations were found among the latent variables; Withdrawal with
Attack Other .63 (p < .001), with Attack Self .82 (p < .001), with Avoidance .23
(p < .001), Attack Other with Attack Self .53 (p < .001), with Avoidance .44
(p < .001); Attack Self with Avoidance .11 (p < .05).

ITEM INDEPENDENCE
Although the CoSS is not an ipsative measure the format could have led
subjects to respond as if it were. If the subjects responded to the CoSS as an
ipsative measure then we would expect (Hicks, 1970): 1) the 4 responses within
each of the 12 stems to be negatively correlated with each other, 2) the 4 scales to
be negatively correlated with each other, and 3) the mean correlation of all scales
with any variable to approach zero. This was not the case. First, the average
correlations within each of the twelve stems were always positive, ranging from
.025 to .249, with a median of .178. The variances ranged from .001 to .069,
with a median of .042. Second, the mean correlation among the scales was .39.
228 COMPASS OF SHAME SCALE

Third, the mean correlation between the scales and other variables is not zero,
as can be seen in Table 1.

TABLE 1
ORDERING OF COSS SCALES BY PREDICTED MAGNITUDE OF CORRELATIONS VERSUS OBSERVED
CORRELATIONS

Validity Construct WD AS AO AV

Shame (ISS) 1a 1a 3b 4b
.71*** .72*** .31* .16
Self-Esteem (ISS) -1a -1a,b -3b, c 0c
-.55*** -.37** -.21 .10
Global Severity Index (SCL-90-R) 2b 1a 3b 0c
.40*** .52*** .34*** .14
Positive Symptom Distress Index (SCL-90-R) 2b 1a 3b 0c
.33*** .54*** .22** .05
Depression (SCL-90-R) 2b 1a 3c 0d
.44*** .57*** .29*** .11
Interpersonal Sensitivity (SCL-90-R) 1a 1a 3b 4c
.52*** .54*** .38*** .14
Hostility-Anger (SCL-90-R) 0b 2a 1a 0b
.15 .30*** .32*** .13
Anger (AOS) 0b 2b 1a 0b
.21 .25* .63*** .11
Distancing (WCQ) -2b -2b -2b 1a
-.20 -.31* -.30* .28*
Escape-Avoidance (WCQ) 1a 0a,b 1a, b 0b
.38** .39** .31* .04
Accepting Responsibility (WCQ) 0a,b 1a 0a,b 0b
.24* .26* .08 -.10
Confrontive Coping (WCQ) 0a 0a 1a 0a
.10 .07 .28* .20
Self-Controlling (WCQ) 0 0 0 0
.18 .14 .01 .19
Seeking Social Support (WCQ) 0 0 0 0
.16 .18 .08 -.11
Planful Problem Solving (WCQ) 0 0 0 0
-.03 .00 -.05 .09
Positive Reappraisal (WCQ) 0 0 0 0
.06 .09 .00 .03

Note: 1 indicates largest predicted magnitude. 0 indicates no correlation predicted. Row predictions
with different subscripts indicate the observed correlations differ significantly at p < .05 (Meng
et al., 1992). WD – Withdrawal; AS – Attack Self; AO – Attack Other; AV – Avoidance. ISS –
Internalized Shame Scale; n = 50. SCL-90-R – Symptom Checklist; n = 150. AOS – Attack Other
Scale; n = 47. WCQ – Ways of Coping Questionnaire; n = 47.
*p < .05. **p < .01. ***p < .001. All one-tailed.
 COMPASS OF SHAME SCALE 229
RELIABILITY
Internal consistency was examined for the four a priori scales. Cronbach’s
alphas were acceptable (Withdrawal, .89; Attack Other, .85; Attack Self, .91;
Avoidance, .74). Three-week test-retest reliabilities for 27 participants were:
Withdrawal, .75; Attack Other, .85; Attack Self, .81; Avoidance, .75.

MEAN DIFFERENCES
The four scales differed in their frequency of use (Attack Self, M = 24.20,
SD = 9.60; Withdrawal, M = 21.89, SD = 9.28; Avoidance, M = 22.59, SD =
6.25; Attack Other, M = 14.93, SD = 7.37). Attack Self was the most commonly
endorsed and Attack Other was endorsed the least.

CONVERGENT VALIDITY WITH CRITERION MEASURES

To investigate convergent validity Pearson correlations were computed
among scores from the individual CoSS scales and scores from the instruments
described previously. These correlations and related hypotheses are displayed in
Table 1. Tests for the significance of differences between correlated correlation
coefficients were computed (Meng, Rosenthal, & Rubin, 1992); these results are
also displayed in Table 1.
Shame (Internalized Shame Scale – ISS) Given the “internalized” (trait) nature
of the ISS, strong correlations with the Withdrawal (r = .71) and Attack Self
(r = .72) scales were expected and obtained; both correlations were significantly
greater than the ISS correlations with the Attack Other (r = .31) and Avoidance
(r = .16) scales. Only the Avoidance correlation was nonsignificant.
Self-esteem (Internalized Shame Scale – ISS-SE) Due to the consciousness
and internalization of shame shared by the Withdrawal and Attack Self scales,
moderately strong negative correlations with the ISS Self-Esteem scale were
predicted and obtained (Table 1; r = -.55 and -.37, respectively). However, the
Attack Other correlation was nonsignificant. The magnitudes of these correlations
with self-esteem for all four CoSS scales were similar to the CoSS correlations
reported by Yelsma, Brown and Elison in their study (2002) of self-esteem. As
expected, greater consciousness and internalization of shame were, in general,
reflected in lower self-esteem. The significant differences between correlations
suggest externalization via Attack Other, and minimization via Avoidance, may
act to protect self-esteem in comparison to Withdrawal and Attack Self.
Psychological Symptoms (Symptom Check List – SCL-90-R) In general,
Compass of Shame styles were predicted to relate to psychological symptoms
according to the degree to which they involve consciousness and internalization
of shame, with the caveat that Attack Self was expected to have the strongest
relationship due to the intrapunitive nature of these scripts. These were the
hypotheses proposed for the GSI, the PSDI, and Depression. The hypotheses
230 COMPASS OF SHAME SCALE

were supported for all three measures; however the correlations with the GSI and
PSDI were not significantly higher for Withdrawal than for Attack Other.
The Interpersonal Sensitivity scale was hypothesized to follow the general
hypothesis with the exception that Withdrawal and Attack Self were not expected
to differ. The results supported these hypotheses; however, the Avoidance
correlation was nonsignificant. It was hypothesized that Hostility would
primarily be related to Attack Other and – to a lesser degree – to Attack Self, but
related to neither Avoidance nor Withdrawal. Again, the results supported these
hypotheses; however the correlation with Attack Other was not significantly
higher than for Attack Self. In summary, Compass of Shame styles demonstrated
a differentiated pattern of correlations with psychological symptoms.
Anger (Attack Other Scale – AOS) Due to the anger inherent in the CoSS
Attack Other Subscale and, to a lesser extent the Attack Self scale, they were
predicted to correlate with measures of trait anger. Withdrawal and Avoidance
were predicted to be unrelated to trait anger. The correlations with the AOS
supported these hypotheses; however the correlation with Attack Self was not
significantly higher than the correlations with Withdrawal and Avoidance.
Coping (Ways of Coping Questionnaire) Of the eight Ways of Coping scales
(Folkman & Lazarus, 1988), four scales (Distancing, Escape-Avoidance,
Accepting Responsibility, Confrontive Coping) were expected to be significantly
correlated with some CoSS scales. These scales are largely emotion-focused in
ways similar to specific CoSS scales.
The Distancing scale of the Ways of Coping Questionnaire “describes cognitive
efforts to detach oneself and to minimize the significance of the situation”
(Folkman & Lazarus, 1988, p.11), thus representing emotional distancing.
Distancing is a feature of the Compass of Shame Avoidance style and in contrast
to the emotional nature of the Attack Self, Withdrawal, and Attack Other styles.
The CoSS correlations with the Distancing scale supported these hypotheses;
however, the Withdrawal correlation, while negative, was nonsignificant.
The items in the Escape-Avoidance scale of the Ways of Coping Questionnaire
acknowledge a problem and associated distress. They describe attempts to
escape from the problem such as “avoided being with people in general” or “took
it out on other people,” behaviors exhibited by those who employ the Compass
of Shame Withdrawal and Attack Other styles. Although the Ways of Coping
Questionnaire and the CoSS both contain scales labeled “Avoidance,” their
meanings are very different. CoSS Avoidance represents emotional detachment
or disavowal; in contrast, the Ways of Coping Questionnaire Escape-Avoidance
items acknowledge the emotionality of the situation and the resulting need
to withdraw. Therefore, no correlation between the two “Avoidance” scales
was expected. The observed correlations supported these hypotheses, with the
exception of the unexpected relationship between Escape-Avoidance and Attack
 COMPASS OF SHAME SCALE 231
Self, which did not differ significantly from the correlations with Withdrawal
and Attack Other. The correlation with Attack Self may have been due to the
acknowledgement of negative emotion shared by both scales.
Accepting Responsibility contains themes of internalization, self-criticism,
and deference, which resulted in significant correlations with Attack Self and
Withdrawal. Confrontive Coping includes hostility and anger, which, as predicted,
resulted in a significant correlation with Attack Other. The remaining scales
describe predominately problem-focused behaviors. Therefore, relationships
between these four scales and CoSS scales were not predicted. None of these
scales demonstrated significant correlations with the CoSS scales.

MALE/FEMALE DIFFERENCES
Male/Female differences were seen in three of the four scales. Men exhibited
more avoidance than did women (t = 1.98, d =.24) and women exhibited more
Attack Self (t = 5.04, d = .61) and more Withdrawal (t = 5.07, d = .61). While
women also exhibited more Attack Other the results were neither statistically
significant nor large (t =1.53, d = .19).

CONCLUSION

Its psychometric properties indicate that the CoSS is a promising instrument for
the assessment of shame-coping styles. It has the expected internal structure. The
four scales have acceptable to high internal consistency. The scales demonstrate
the expected pattern of correlations with criterion variables. Withdrawal and
Attack Self share characteristic features of acceptance and internalization of
shame. Illustrating the intrapunitive nature of Attack Self scripts, the Attack Self
scale demonstrates the strongest relationships with psychological symptoms in
terms of the Global Severity Index, the Positive Symptom Distress Index, and
Depression. Attack Other scripts serve to externalize or redirect shame. The
Attack Other scale is differentiated by greater hostility and anger and has been
associated with psychopathy (Campbell & Elison, 2005). Avoidance scripts
serve to disavow or minimize shame. Fewer psychological symptoms and the
use of Distancing as a coping strategy characterize Avoidance. The internal
structure and the numerous validity measures support the Compass of Shame
as a theoretical model and support the CoSS as a tool which accurately assesses
this model.

CORRELATIONS AMONG FACTORS

With one exception the correlations among factors appear to reflect the
features shared by specific pairs of poles: acceptance shared by Attack Self and
Withdrawal, minimization shared by Attack Other and Avoidance, and anger
232 COMPASS OF SHAME SCALE

shared by Attack Self and Attack Other. As expected, Attack Self and Withdrawal
have the highest correlation among any pair of factors. These poles share the
conscious and internalized nature of the shame experience; shame is accepted as
valid and there is a willingness to acknowledge feeling bad (Nathanson, 1992). In
contrast, the Attack Other and Avoidance poles share the minimization of shame,
which accounts for their moderate correlation. A moderate correlation is seen
between both Attack poles. They share the anger characteristic and one might
expect the target of the anger to vacillate between self and others (Nathanson,
1992). As expected, the pairs Attack Self and Avoidance and Withdrawal and
Avoidance exhibit low correlations.
The one exception is the unexpectedly high correlation found between Attack
Other and Withdrawal. This correlation may represent passive-aggressive
behavior, passive behaviors of the Withdrawal pole and aggressive behaviors of
the Attack Other pole. Evidence for a connection between a desire to withdraw
and anger was found by Wicker, Payne, and Morgan (1983) and Tangney, Miller,
and Flicker (1992). Empirical support for this interpretation comes from Cook’s
(1996) work with his Internalized Shame Scale and his Attack Other Scale (see
Measures section). Cook’s results include correlations between the MCMI-II
Passive Aggressive scale and the ISS of .65, and with the AOS of .64, both
significant at p < .001.
The CoSS scales are not orthogonal; nor were they meant to be. The
correlations among the factors suggest that individuals frequently employ more
than one script, but it does not mean the scales are redundant. Folkman and
Lazarus (1988) note that an important question in coping research is the temporal
unfolding of coping processes. CoSS scales may be correlated with each other
and with similar magnitudes to other variables because one CoSS style often
follows another. In addition, Folkman and Lazarus note that situational factors,
such as the type of stressor and whether the situation is changeable, affect
choice of coping styles. Because the CoSS focuses specifically on coping with
shame, situational factors are less variable in comparison to the Ways of Coping
Questionnaire where a wide variety of stressors are involved. Thus, CoSS scales
may show less discrimination than measures of coping with stressors in general.
Nonetheless, even in the case of the most highly correlated scales, Attack Self
and Withdrawal, over 16% of the participants differ on the two scales by more
than one standard deviation. Studies of their temporal relationships or with
criterion variables assessing the differences suggested by Nathanson (e.g.,
conformity, shyness) may better reveal discrimination between Withdrawal and
Attack Self.
Convergent Validity Significant correlations between CoSS scales and criterion
measures are evident in almost all cases where they were predicted (28 of 32).
As predicted, the intrapunitive nature of Attack Self scripts acts to magnify
 COMPASS OF SHAME SCALE 233
the destructive aspects (e.g., depression, symptom distress index) of shame;
Withdrawal and Attack Self scripts are characterized by internalization (e.g.,
internalized shame, lowered self-esteem); Attack Other scripts involve exter-
nalization (e.g., hostility, anger); Avoidance acts through emotional minimization
(e.g., distancing).
Discriminant Validity When validating a scale composed of multiple subscales,
such as the CoSS, discriminant validity encompasses two separate issues.
The first issue is discrimination in general; in other words, are CoSS scales
uncorrelated with measures for which no relationship is expected? The second
issue is discrimination among CoSS scales; in other words, do the magnitudes
of the correlations between CoSS scales and a given criterion measure differ?
In terms of general discrimination, CoSS scales are not significantly correlated
with measures of problem-focused coping, with one anticipated exception (i.e.,
Attack Other – Confrontive Coping). Nor are they significantly correlated with
Seeking Social Support, an emotion-focused style unlike any CoSS scale. Thus,
in our limited test, the results support general discriminant validity.
In terms of discrimination among CoSS scales, the magnitudes of the
correlations usually differ across scales as predicted (54 of 56 hypotheses);
in many cases these differences are significant (40 of 56). This differentiated
pattern is important because Compass of Shame styles are believed to mediate
the relationships between the experience of shame and other variables, with
different relationships predicted for each of the four styles (Nathanson, 1992).
However, a number of important exceptions are evident. While the magnitudes
of correlations between Attack Other and criterion variables are lower than
for Withdrawal and Attack Self as predicted, in several cases these differences
are not significant. More importantly, the magnitudes of correlations between
Withdrawal and Attack Self and the majority of criterion measures do not differ
significantly. While many of these similarities were anticipated, the hypothesized
differences between Withdrawal and Attack Self are significant in only four of
seven cases.

FUTURE APPLICATIONS
Nathanson (1992) hypothesized that Compass of Shame styles are differentially
related to variables associated with shame (e.g., depression, hostility), as would
be the case if shame coping mediated the relationships between the experience
of shame and these variables. While mediation was not explored in the
present study, the results support the hypothesis of differentiated relationships
with regard to a variety of constructs: internalization of shame, self-esteem,
psychological symptoms, anger, and other coping styles. Similarly, other
studies have supported this hypothesis with regard to self-esteem (Yelsma
et al., 2002), psychopathy (Campbell & Elison, 2005), anger, and a general
234 COMPASS OF SHAME SCALE

measure of psychological symptoms (Elison et al., 2006). Recent reviews of

the shame and guilt literature encourage investigations of when, and under what
conditions, these emotions are adaptive versus maladaptive (Bybee & Quiles,
1998; Ferguson & Stegge, 1998; Tangney & Dearing, 2002). The CoSS and
results reported here provide a starting point from which these questions may be
addressed. Each of the four CoSS scales assesses a different approach to shame
coping and each approach may be adaptive or maladaptive depending on the
context in which it is employed.
From the results presented here one might infer that Avoidance is a more
effective or adaptive shame-focused coping style than the others. This conclusion
depends on one’s definition of effective and adaptive. Certainly Avoidance
exhibits weaker associations with psychological symptoms as measured in this
study. However, Avoidance may have other associated costs, which reduce its
effectiveness. Nathanson’s theory (1992) suggests that addictions and narcissism
may be among the costs of Avoidance. Furthermore, the denial aspect of
Avoidance coping may serve to attenuate correlations. People who employ
Avoidance would be less likely to be aware of psychological symptoms, even
when present, a limitation of self-report studies. A more complete assessment
of Avoidance may require ratings of participants by other sources and measures
of other attributes, such as addictions, narcissism, and quality of relationships.
Therefore, conclusions about which styles are more effective, adaptive, or
healthy should be framed in a broader context. These are issues to be pursued
in future studies.
There has been an increased focus on the role of shame in the development of
self, self-esteem, and psychopathology. The tragedy at Columbine High School
and more recent school shootings have resulted in increased awareness of the
role of shame in violence (Elison & Harter, 2004; Harter, Low, & Whitesell,
2003; Nathanson, 2003); focus has switched from the bullies to the bullied.
The relevance of the Compass of Shame model to the development of self,
self-esteem, psychopathology, and violence suggests its value in research,
psychotherapy, and interventions. Indeed, the current study provides empirical
support for the major theoretical prediction of this model, specifically a
differentiated pattern of associations between shame-coping styles and related
constructs. The way in which one copes with, or defends against, shame truly is
an important aspect of the shame experience. The reliability, structural validity,
and criterion-related validity of the CoSS support its use as a measure of the
underlying Compass of Shame model.

REFERENCES

Bybee, J., & Quiles, Z. N. (1998). Guilt and mental health. In J. Bybee (Ed.), Guilt and children (pp.
269-291). New York: Academic Press.
 COMPASS OF SHAME SCALE 235
Campbell, J. S., & Elison, J. (2005). Shame coping styles and psychopathic personality traits.
Journal of Personality Assessment, 84, 96-104.
Cook, D. R. (1996). [Reliability and validity of Attack Other Scale (AOS)]. Unpublished raw data.
Cook, D. R. (2001). Internalized Shame Scale: Technical manual. Toronto: Multi-Health Systems.
Derogatis, L. R. (1983). SCL-90-R manual-II. Towson, MD: Clinical Psychometric Research.
Elison, J., & Harter, S. (2004, April). Anger and violent ideation as correlates of shame,
embarrassment, and humiliation. Paper presented at the annual meeting of the Rocky Mountain
Psychological Association, Reno, NV.
Elison, J., Pulos, S., & Lennon, R. (2006). Shame-focused coping: An empirical study of the
Compass of Shame. Social Behavior and Personality, 34 (2), 161-168.
Ferguson, T. J., & Stegge, H. (1998). Measuring guilt in children: A rose by any other name still has
thorns. In J. Bybee (Ed.), Guilt and children (pp. 19-74). New York: Academic Press.
Folkman, S., & Lazarus, R. S. (1988). Manual for the Ways of Coping Questionnaire. Palo Alto:
Consulting Psychologists Press.
Folkman, S., Lazarus, R. S., Dunkel-Schetter, C., DeLongis, A., & Gruen, R. J. (1986). Dynamics
of a stressful encounter: Cognitive appraisal, coping, and encounter outcomes. Journal of
Personality and Social Psychology, 50, 992-1003.
Harder, D. W. (1995). Shame and guilt assessment, and relationships of shame and guilt-proneness
to psychopathology. In J. P. Tangney & K. W. Fischer (Eds.), Self-conscious emotions: The
psychology of shame, guilt, embarrassment, and pride (pp. 368-392). New York: Guilford
Press.
Harder, D. W., & Zalma, A. (1990). Two promising shame and guilt scales: A construct validity
comparison. Journal of Personality Assessment, 55, 729-745.
Harter, S., Low, S. M., & Whitesell, N. R. (2003). What have we learned from Columbine: The
impact of the self-system on suicidal and violent ideation among adolescents. Journal of School
Violence, 2, 3-26.
Hicks, L. E. (1970). Some properties of ipsative, normative, and forced-choice normative measures.
Psychological Bulletin, 74, 167-184.
Horowitz, M. J. (1979). States of mind. New York: Plenum Press.
Lazarus, R. S. (1999). Stress and emotion: A new synthesis. New York: Springer.
Lewis, H. B. (1971). Shame and guilt in neurosis. New York: International Universities Press.
Lewis, H. B. (1987). The role of shame in symptom formation. Hillsdale, NJ: Erlbaum.
McDonald, R. P., & Ho, M. R. (2002). Principles and practice in reporting structural equation
analyses. Psychological Methods, 7, 64–82.
Meng, X., Rosenthal, R., & Rubin, D. B. (1992). Comparing correlated correlation coefficients.
Psychological Bulletin, 111, 172-175.
Nathanson, D. L. (1992). Shame and pride. New York: Norton.
Nathanson, D. L. (1994). Shame, compassion and the “borderline” personality. Psychiatric Clinics
of North America, 17, 785-810.
Nathanson, D. L. (Producer/Writer). (2003). Managing shame, preventing violence: A call to our
clergy [Digital Video Disk]. (Available from the Silvan S. Tomkins Institute, 255 South 17th
Street (Suite 2403), Philadelphia, PA 19103-6224)
Parker, S. T. (1998). A social selection model for the evolution and adaptive significance of self-
conscious emotions. In M. Ferrari & R. J. Sternberg (Eds.), Self-awareness: Its nature and
development (pp. 108-134). New York: Guilford Press.
Tangney, J. P., & Dearing, R. L. (2002). Shame and guilt. New York: Guilford Press.
Tangney, J. P., Miller, R. S., & Flicker, L. (1992, August). A quantitative analysis of shame and
embarrassment. Poster presented at the meeting of the American Psychological Association,
Washington, DC.
236 COMPASS OF SHAME SCALE

Tangney, J. P., Wagner, P. E., & Gramzow, R. (1989). The Test of Self-Conscious Affect. Fairfax, VA:
George Mason University.
Tomkins, S. S. (1962 -1991). Affect/imagery/consciousness. Vols. 1-3. New York: Springer.
Wei, S., Lennon, R., & Pulos, S. (2005). The relation of parenting to adolescents’ coping styles.
Manuscript submitted for publication.
Wicker, F. W., Payne, G. C., & Morgan, R. D. (1983). Participant descriptions of guilt and shame.
Motivation and Emotion, 7, 25-39.
Yelsma, P., Brown, N. M., & Elison, J. (2002). Shame-focused coping styles and their associations
with self-esteem. Psychological Reports, 90, 1179-1189.
 COMPASS OF SHAME SCALE 237
APPENDIX
COSS
(VERSION 3)
Directions: Below is a list of statements describing situations you may experience from time to time.
Following each situation are four statements describing possible reactions to the situation. Read each
statement carefully and circle the number to the left of the item that indicates the frequency with
which you find yourself reacting in that way. Use the scale below. Please respond to all four items
for each situation.

SCALE

0 1 2 3 4

NEVER SELDOM SOMETIMES OFTEN ALMOST ALWAYS

SCALE
A. When an activity makes me feel like my strength or skill is inferior:
0 1 2 3 4 1. I act as if it isn’t so. (AV)
0 1 2 3 4 2. I get mad at myself for not being good enough. (AS)
0 1 2 3 4 3. I withdraw from the activity. (WD)
0 1 2 3 4 4. I get irritated with other people. (AO)

B. In competitive situations where I compare myself with others:

0 1 2 3 4 5. I criticize myself. (AS)
0 1 2 3 4 6. I try not to be noticed. (WD)
0 1 2 3 4 7. I feel ill will toward the others. (AO)
0 1 2 3 4 8. I exaggerate my accomplishments. (AV)

C. In situations where I feel insecure or doubt myself:

0 1 2 3 4 9. I shrink away from others. (WD)
0 1 2 3 4 10. I feel others are to blame for making me feel that way. (AO)
0 1 2 3 4 11. I act more confident than I am. (AV)
0 1 2 3 4 12. I feel irritated with myself. (AS)

D. At times when I am unhappy with how I look:

0 1 2 3 4 13. I take it out on other people. (AO)
0 1 2 3 4 14. I pretend I don’t care. (AV)
0 1 2 3 4 15. I feel annoyed at myself. (AS)
0 1 2 3 4 16. I keep away from other people. (WD)

E. When I make an embarrassing mistake in public:

0 1 2 3 4 17. I hide my embarrassment with a joke. (AV)
0 1 2 3 4 18. I feel like kicking myself. (AS)
0 1 2 3 4 19. I wish I could become invisible. (WD)
0 1 2 3 4 20. I feel annoyed at people for noticing. (AO)
238 COMPASS OF SHAME SCALE

F. When I feel lonely or left out:

0 1 2 3 4 21. I blame myself. (AS)
0 1 2 3 4 22. I pull away from others. (WD)
0 1 2 3 4 23. I blame other people. (AO)
0 1 2 3 4 24. I don’t let it show. (AV)

G. When I feel others think poorly of me:

0 1 2 3 4 25. I want to escape their view. (WD)
0 1 2 3 4 26. I want to point out their faults. (AO)
0 1 2 3 4 27. I deny there is any reason for me to feel bad. (AV)
0 1 2 3 4 28. I dwell on my shortcomings. (AS)

H. When I think I have disappointed other people:

0 1 2 3 4 29. I get mad at them for expecting so much from me. (AO)
0 1 2 3 4 30. I cover my feelings with a joke. (AV)
0 1 2 3 4 31. I get down on myself. (AS)
0 1 2 3 4 32. I remove myself from the situation. (WD)

I. When I feel rejected by someone:

0 1 2 3 4 33. I soothe myself with distractions. (AV)
0 1 2 3 4 34. I brood over my flaws. (AS)
0 1 2 3 4 35. I avoid them. (WD)
0 1 2 3 4 36. I get angry with them. (AO)

J. When other people point out my faults:

0 1 2 3 4 37. I feel like I can’t do anything right. (AS)
0 1 2 3 4 38. I want to run away. (WD)
0 1 2 3 4 39. I point out their faults. (AO)
0 1 2 3 4 40. I refuse to acknowledge those faults. (AV)

K. When I feel humiliated:

0 1 2 3 4 41. I isolate myself from other people. (WD)
0 1 2 3 4 42. I get mad at people for making me feel this way. (AO)
0 1 2 3 4 43. I cover up the humiliation by keeping busy. (AV)
0 1 2 3 4 44. I get angry with myself. (AS)

L. When I feel guilty:

0 1 2 3 4 45. I push the feeling back on those who make me feel this way. (AO)
0 1 2 3 4 46. I disown the feeling. (AV)
0 1 2 3 4 47. I put myself down. (AS)
0 1 2 3 4 48. I want to disappear. (WD)