Content uploaded by Mark Blais
Author content
All content in this area was uploaded by Mark Blais on Sep 21, 2017
Content may be subject to copyright.
Vol. 73, No. 3 (Summer 2009) 203
A previous version of this article was presented at the 2001 Midwinter Meeting of the
Society for Personality Assessment, Philadelphia, PA.
Dr. Baity is Assistant Professor at Alliant International University California School
of Professional Psychology – Sacramento. Dr. Blais is Director, Psychological Evalua-
tion and Research Laboratory at Massachusetts General Hospital / Harvard Medical
School. Dr. Hilsenroth is Associate Professor at Adelphi University. Dr. Fowler is Di-
rector of Research at Austen Riggs Center. Dr. Padawer is an Organizational Consul-
tant at Salt Lake City, Utah.
Correspondence may be sent to Dr. Baity at CSPP, 2030 W El Camino Avenue, Sacra-
mento, CA 95833; e-mail: mbaity@alliant.edu.
Baity et al.
Rorschach and Self-Mutilation
Self-mutilation, severity of borderline
psychopathology, and the Rorschach
Matthew R. Baity, PhD
Mark A. Blais, PsyD
Mark J. Hilsenroth, PhD
J. Christopher Fowler, PhD
Justin R. Padawer, PhD
The authors explore borderline pathology on a continuum of func-
tioning. Rorschach variables relating to (1) aggression, (2) dependen-
cy, (3) object relations, (4) defenses, and (5) boundary disturbance
were measured across a nonclinical (NC) and two clinical (borderline
patients without self-mutilative behavior = N-BPD, and borderline
patients with self-mutilative behaviors = SM-BPD) groups. Results
demonstrated good discriminate ability (87%) between clinical and
nonclinical protocols. Comparisons between N-BPD and SM-BPD
groups revealed overall greater pathological scores for the SM-BPD
group, specifically in dependency scores. Convergence with other
studies and implications for future clinical and empirical work are
discussed. (Bulletin of the Menninger Clinic, 73 [3], 203-225]
Although borderline personality disorder (BPD) has been part of
the official psychiatric nomenclature for over 30 years, the con-
cepts of borderline personality and borderline psychopathology
did not begin with the DSM (Gunderson, 2001). The earlier psy-
choanalytic concepts of borderline states (Knight, 1953) and then
borderline personality organization (BPO; Kernberg, 1975) laid the
foundation for the DSM BPD. In writing about borderline states,
Baity et al.
204 Bulletin of the Menninger Clinic
Knight highlighted the fact that these patients evidenced severe,
nonspecific ego weaknesses, including brief periods of poor reality
contact and a propensity for regression in unstructured settings,
despite initially appearing to be more psychologically intact. Kern-
berg (1970, 1975) advanced our understanding of these patients
further by developing a complex hierarchical psychoanalytic model
of personality development and organization. Within this model,
patients are placed on a continuum of functioning ranging from
psychosis to neurosis. Core features of individuals organized at a
borderline personality organization (BPO) include identity diffu-
sion, disturbances in object relations, lapses in reality contact, reli-
ance on primitive defenses such as splitting and projection, and
excessive aggression (Blais and Baity, 2008).
The important distinction between the DSM and the Kernber-
gian approaches to BPD is the use of categorical versus develop-
mental/dimensional approaches to diagnosis. Unlike diagnoses
such as major depressive episode, which require the presence of ei-
ther depressed mood or loss of interest, the DSM BPD criteria have
no identified core features so that two patients meeting criteria for
BPD could look quite different clinically. Conversely, Kernberg ap-
proaches BPD from a developmental standpoint and requires cli-
nicians to assess patients on several functional continuums (e.g.,
defensive structure, quality of relationships). The focus of assess-
ment in Kernberg’s model is on how an individual’s personality is
organized based on select theoretical variables. Albeit broader, this
approach to assessing borderline personality provides the clinician
with richer clinical material than the traditional DSM-IV criteria
(Blais & Baity, 2008). The Rorschach inkblot test is considered an
ideal instrument for assessing the dimensional features of border-
line psychopathology due both to the nature of the Rorschach task,
and to the rich tradition of assessing pathological character styles
with this measure.
Lerner, Sugarman, and Barbour (1985) used the Rorschach to
differentiate the interpersonal boundaries of severely disturbed
clinical groups, a task originated by Rapaport, Gill, and Schafer
(1945). Results from this study showed that the inpatient schizo-
phrenic sample showed a greater tendency to merge two concepts
onto one area of the blot (Self-Other boundaries) than any of the
other groups. Inpatients diagnosed with BPD had significantly more
Inner-Outer boundary disturbances on the Rorschach than either
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 205
a group of inpatient schizophrenics or outpatients with BPD. In
other words, inpatients with BPD had great difficulty distinguish-
ing their actual affective states from their fantasized ones (“Oh my
God, it’s a monster lunging right for me!”). When patients with
BPD require inpatient level of care, Lerner et al. suggest that these
patients’ fantasy life and real life have become so intertwined that
they can no longer negotiate these boundaries on their own. The
findings from this study are important for two reasons: First, they
help characterize the types of boundary disturbances that occur
among severe psychiatric disorders and, second, they provide evi-
dence that diffuse ego boundary responses from the inpatient with
BPD occur along a spectrum of severity.
Blais, Hilsenroth, Fowler, and Conboy (1999) used Rorschach
variables to explore the DSM-IV BPD criteria set. Earlier studies of
the content validity (Blais, Hilsenroth, & Castlebury, 1997) and hi-
erarchical structure (Blais, Hilsenroth, & Fowler, 1999) of DSM-IV
BPD revealed that the criteria seemed to cluster into core features
of the disorder. The domains thought to be central descriptors of
BPD include difficulties with identity, affective instability, and un-
stable interpersonal relationships. Blais, Hilsenroth, Fowler, et al.
(1999) examined these core features further by using Rorschachs
taken from outpatients reliably diagnosed with DSM-IV Cluster B
personality disorders. The authors selected Rorschach scales that
had theoretical relevance to the core BPD features. These variables
were Holt’s aggression scores (1977), the Rorschach Oral Depen-
dency Scale (ROD; Masling, Rabie, & Blondheim, 1967), the Mu-
tuality of Autonomy Scale (MOAS; Urist, 1977), and the Lerner
Defense Scales (LDS; Lerner, 1991). The profile that emerged for
BPD from this study was that, when compared to patients with
other Cluster B disorders, patients with BPD had more raw aggres-
sive impulses, greater disruptions in object relations, and greater
use of low-level defenses (Blais, Hilsenroth, Fowler, et al., 1999).
In addition to being the first to examine the relationship between
the Rorschach and DSM-IV BPD criteria, this study demonstrated
the importance of using the Rorschach to aid in the differential
diagnosis of BPD from other Cluster B disorders.
Perhaps one of the clearest examples of how borderline pathol-
ogy extends along a continuum of severity is to examine those
BPD patients who engage in self-mutilative behaviors (eg., cutting,
burning, scratching) versus those who do not self-mutilate. Several
Baity et al.
206 Bulletin of the Menninger Clinic
psychoanalytically informed theories have emerged over the years
to help explain the etiology and function of self-mutilation (Doc-
tors, 1981; Gunderson, 2001; Kernberg, 1975, 1984; McMillian,
1994; Stolorow & Lachmann, 1980; see also Fowler & Hilsenroth,
1999). A consistent theme across these writings on self-mutilation
and BPD is that these individuals have a very primitive and fragile
internal world, which makes them highly sensitive to even minor
shifts in their emotional states. When affective shifts occur, they
can be highly disorganizing and usually occur in the context of
important relationships, either feeling too close or too distant. In
general, real or imagined shifts in relationships can cause a great
deal of internal tension to the extent that individuals with BPD
can quickly feel either completely abandoned by others or wholly
enmeshed with them (McMillian, 1994). Interpersonal disruptions
cause a great deal of emotional turmoil that often results in frenetic
searches for control, or at least relief. The ensuing self-mutilation is
seen as a method of regaining that equilibrium.
In summarizing the theories on self-mutilation and borderline
pathology Fowler, Hilsenroth, and Nolan (2000) identified five
areas of psychological functioning that related to this behavior:
(1) intense, unmodulated aggressive affects, (2) high dependency,
(3) malevolent object representations, (4) frequent use of primitive
defenses (i.e., splitting, devaluation, idealization), and (5) frequent
boundary disturbances between themselves and others. These areas
are clearly related to the core features based on DSM-IV criteria (de-
scribed earlier), which included difficulties with identity, affective
instability, and unstable interpersonal relationships (Blais, Hilsen-
roth, Fowler, et al., 1999). Fowler and colleagues (2000) hypothe-
sized that borderline patients who exhibit self-mutilating behaviors
would appear more pathological on these five common areas than
borderline patients who do not self-mutilate. This study was one of
the most comprehensive attempts to empirically validate differenc-
es in borderline pathology using dynamically informed Rorschach
scales. Results from this study showed that self-mutilating border-
line patients had significantly greater disturbances in all but one of
these five core characteristics (no differences in dependency scores
between groups; Fowler et al., 2000). That is to say, borderline pa-
tients with self-harming behaviors provided responses laden with
more intense aggression, malevolent relationship expectations,
primitive defenses, and boundary violations than patients without
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 207
a self-mutilative history. This study is important in its description
of the internal world of patients who self-mutilate, as well as pro-
viding strong evidence for the existence of borderline pathology as
a dimensional variable.
Based on a review of the current literature, it appears that em-
pirical investigations have been able to distinguish varying levels of
borderline pathology on specific psychological domains that can
be assessed with the Rorschach. Research has also suggested that
borderline pathology exists on a continuum where patients who
self-mutilate generally appear more disturbed on the Rorschach
than their non-self-mutilating counterparts (Fowler & Hilsenroth,
1999; Fowler et al., 2000). However, much of the research thus far
has focused primarily on mean score differences between groups
that place some limitations on conclusions that can be drawn about
the data. The current study will attempt to expand on previous
findings by comparing the Rorschach scales (BDS, Holt’s aggres-
sion scores, LDS, MOAS, and ROD) that have both theoretical and
empirical ties to the five core features of self-mutilating borderline
patients across nonclinical (NC), non-self-mutilating (N-BPD), and
self-mutilating patients diagnosed with BPD. The current study is
distinctive in that the Rorschach scales will be used in regression
analyses to try to predict not only total BPD criteria but also self-
mutilation. Based on previous findings, it is expected that distinct
levels of pathology will be established, with the nonclinical group
appearing the healthiest, the self-mutilating BPD group (SM-BPD)
having the most pathological scores, and the non-self-mutilating
group (N-BPD) being somewhere in between. Because few other
studies have reported similar analyses, it is difficult to hypothesize
which Rorschach scale or scales will show the greatest utility in
predicting group memberships.
Methods
Participants
A total of 100 Rorschach protocols (50 clinical and 50 nonclin-
ical) were used for the current study. Demographics for the full
sample included 32 males and 68 females with a mean age of 27
(SD 8.4) years and an average of 14 (SD 1.8) years of education.
Seventy-three participants were single, 11 were married, and 15
Baity et al.
208 Bulletin of the Menninger Clinic
were divorced. The nonclinical (NC) subsample consisted of Ror-
schachs administered to 50 (25 male and 25 female) undergraduate
psychology majors who received course credit for their participa-
tion. The average age for the NC sample was 22.6 (SD 5.3) years
with 14.8 (SD 1.1) years of education. Forty-six individuals were
single (92%) and 8 were married (8%). The clinical subsample was
compiled from three different studies (Blais, Hilsenroth, & Fowler,
1999; and Blais, Hilsenroth, Fowler, et al., 1999; Fowler, Hilsen-
roth, and Handler, 1998) and included 50 Rorschach protocols of
patients carefully diagnosed with BPD. These patients were being
seen in a variety of settings, including a university-based outpatient
clinic, a short-term inpatient facility, and a hospital-based outpa-
tient clinic. To ensure diagnostic accuracy, two studies report their
interrater reliability statistics for BPD criteria (n = 19 from Blais,
Hilsenroth, & Fowler, 1999, and n = 25 from Blais, Hilsenroth,
Fowler, et al., 1999). The resulting kappa coefficients were .85 and
.80, respectively, which is considered to be in excellent agreement
(Fleiss, 1981). In order to help manage the length of this article
and reduce the likelihood of confusion, readers are encouraged to
reference the original studies for an exact report of the data collec-
tion and reliability procedures for participants used in this study.
Based on interview and medical exam data at intake, 15 patients
were identified as having a recent history of self-mutilative behav-
ior (SM-BPD) while 35 patients (N-BPD) showed no indication of
recent self-harming behaviors within at least the past 6 months.
The total BPD sample consisted of 7 males and 43 females with
a mean age of 31 (SD 9.0) and 13.7 (SD 2.4) years of education.
Twenty-seven of the patients in the BPD sample were single (55%),
11 were divorced (22%), and 11 were married (22%; data missing
for one participant).
Measures
In taking from research designed to better define thought-disor-
dered responses on the Rorschach (Blatt & Ritzler, 1974; Quinlan
& Harrow, 1974; Rapaport, Gill, & Schafer, 1968), Lerner et al.
(1985) devised the Boundary Disturbance Scale (BDS) as a way
to assess disordered thinking on a developmental continuum. The
degree of disturbance is determined, in part, by the level of reali-
ty-based responses and how intact the forms of objects identified
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 209
on the blot appear to be. The BDS consists of a 6-point weighted
scale within 3 areas (Boundary Laxness, Inner-Outer, Self-Other)
of increasing boundary dysfunction. Boundary Laxness includes
two forms combined in an unrealistic fashion but with distinct
boundaries (e.g., “two chickens playing basketball”). The next lev-
el, Inner-Outer Boundary, are responses that give a sense that the
individual is becoming overly involved in the task either through
the use of affectively charged themes or over inclusive descriptions
that commonly include personal references (e.g., “a bullet coming
right for me”). The Self-Other Boundary violations occur when the
distinction between two ideas in a response becomes fluid, condi-
tional, or completely merged. Responses that include implausible
transparent images (e.g., “you can see his heart beating through his
chest”) or impossible combinations (e.g., “a man with frog’s feet”)
are also included in this category.
Aggressive ideation was assessed using Holt’s (1977) concep-
tualization for scoring primary and secondary process responses
on the Rorschach. Primary process aggression (A1) is identified by
responses that include highly aggressive, usually lethal, sadomas-
ochistic themes (e.g., “a sword stuck in someone’s head and there’s
blood pouring out”). The next level in Holt’s system is the second-
ary process score (A2) and includes responses that represent the
expression of aggression in a more socially appropriate (nonlethal)
manner (e.g., “a sword”, “a tank”). The use of Holt’s system to
score aggression on the Rorschach has demonstrated good reli-
ability (Baity & Hilsenroth, 1999; Fowler, Hilsenroth, & Handler,
Table 1. Interrater Reliability for Psychodynamically-oriented
Rorschach Scales (20 protocols).
Variables OCC
Boundary Disturbance .98
Aggression .98
Lerner Defense Scale .94
Mutuality of Autonomy .95
Rorschach Oral Dependency .95
Note. OCC = Overall Correct Classification (hit rate) formula from Kessel and Zimmerman
(1993).
Baity et al.
210 Bulletin of the Menninger Clinic
1995) and validity (Baity & Hilsenroth, 1999; Blatt & Berman,
1984; Hilsenroth, Hibbard, Nash, & Handler, 1993).
Lerner and Lerner developed their Defense Scales (LDS; Lerner
& Lerner, 1980) based on Kernberg’s (1975) conceptualization of
primitive defensives. The LDS operationalizes the defenses of split-
ting, idealizing, devaluing, denial, and projective identification,
as these are depicted in human, quasi-human, and human detail
Rorschach responses. Each of the specific defenses in the LDS is
rated on a continuum from low (less pathological) to high (more
pathological), reflecting the degree of distortion present. To use the
scale, clinically or in research, the ratings for each specific defense
are summed, yielding a total score for each defense. For example, if
three instances of devaluation were identified and they were rated
at the levels of 3, 3, and 1, the patient would receive a total devalu-
ation score of 7 (3 + 3 + 1 = 7). The defenses of splitting, devalu-
Table 2. Descriptive Statistics of the Rorschach Variables
for the Full Sample (N = 100).
Variables X SD MIN MAX Skewness Kurtosis
BDS
Inner-Outer .92 1.6 0.0 7.0 2.1 4.2
Self-Other .60 1.0 0.0 7.0 3.2 16.0
Aggression
A1 .50 .82 0.0 3.0 1.7 2.3
A2 5.4 3.1 0.0 17.0 1.3 2.3
LDS
Split .23 .66 0.0 4.0 4.1 20.0
Deval 5.9 5.4 0.0 23.0 1.0 .64
Ideal 2.7 3.5 0.0 17.0 1.5 2.8
MOAS
MOAS-H* 5.3 1.7 1.0 7.0 -1.0 -.35
MOAS-PATH* 1.6 1.8 0.0 8.0 1.5 2.2
Dependency
ROD 3.7 2.3 0.0 11.0 .71 .71
Note. NC = Nonclinical sample; Total BPD = Total BPD sample; BDS = Boundary Disturbance
Scale; A1 = Holt’s Primary Process Aggression Score; A2 = Holt’s Secondary Process Aggression
Score; LDS = Lerner Defense Scale; MOAS = Mutuality of Autonomy Scale; MOAS-H = Highest
(most pathological) MOAS score on a given protocol; MOAS-PATH = Total number of MOAS
scores of 5, 6, and 7 for a given protocol; ROD = Rorschach Oral Dependency Scale. *N = 99.
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 211
ation, and idealization have shown utility in differentiating BPD
from other psychiatric diagnoses and were chosen a priori for the
current study.
The MOAS (Urist, 1977) examines the relational content that
appears in Rorschach responses between any two objects, percepts,
or ideas. The primary focus of this scale is to assess the degree
to which relational themes include differentiation among objects
included in the response. This 7-point scale begins with the repre-
sentation of objects as separate entities engaged in a mutual activ-
ity (e.g., “two people clinking wine glasses”; a score of 1 or 2).
The next stage (score of 3 or 4) of the MOAS is scored when the
featured objects show signs of decreasing autonomy and increasing
dependency (e.g., “looks like two people connected at the waist”).
A score of 5, 6, or 7 includes responses in which objects not only
lose independence, but also appear in a relationship that is increas-
ingly predatory (e.g., “this Venus flytrap is about to eat this fly”
would get a score of 7). Research has demonstrated good reliabil-
ity (Blais, Hilsenroth, Fowler et al., 1999; Fowler, Hilsenroth, &
Handler, 1996; Fowler et al., 1998) for this scale, as well as use of
the highest (most pathological) MOAS score (MOAS-H). A com-
posite of the total number of 5, 6, and 7 scores (MOAS-PATH;
Berg, Packer, & Nunno, 1993) for a given protocol is a robust
measure of pathological object relating and will be used in the cur-
rent study.
The ROD (Masling et al., 1967) was developed as a scale to mea-
sure oral or dependency needs based on the content of responses
and can be interpreted at three different points. Low scores suggest
individuals who rigidly defend their need for others by distanc-
ing themselves from close relationships. High ROD scores suggest
those who are overly effusive and pursuant in their need for others.
Scores in the low and high range are considered abnormal, while
midrange scores suggest a more adaptive balance between complete
independence and total reliance on others. Protocols receive one
point for each of the following oral dependent percepts: food and
drinks, food sources, food objects, food providers, passive food
receivers, food organs, supplicants, nurturers, gifts and gift givers,
good luck symbols, oral activity, passivity and helplessness, preg-
nancy, reproductive anatomy, and negation of oral percepts (e.g.,
“a man with no mouth”). Protocols were scored for ROD based
on information gathered from both the free association and inquiry
Baity et al.
212 Bulletin of the Menninger Clinic
phases of the Rorschach task (Bornstein, Hilsenroth, Padawer, &
Fowler, 2000). The reliability and validity of this scale has been
well established in previous literature with numerous populations
(Bornstein, 1996).
Procedure
The data used for the current study were screened prior to inclu-
sion, and any protocol with R <14 and L (level of effort in com-
pleting the task) >1 was omitted from further analyses. The Ror-
schach protocols used in the current study were all administered
by advanced graduate students enrolled in either an APA-approved
clinical psychology training program or a predoctoral internship
according to the procedures outlined by Exner (1993). At the time
of the original data collection, 20 protocols were randomly select-
ed and scored by the third and fourth authors for the purpose of
interrater reliability (Weiner, 1991). The raters also independently
scored the dynamically oriented scales used in the current study
during this time. All raters (as well as original scorers) were blind
to diagnosis and/or whether the protocol being scored was admin-
istered to a patient or a student. Table 1 shows very high overall
correct classification rates (OCC; Kessel & Zimmerman, 1993) for
each scale used in the current study, suggesting strong interrater
reliability.
Results
Table 2 shows the descriptive statistics for the Rorschach variables
used in this study. The clinical and nonclinical groups were com-
bined to evaluate the distribution of scores across the entire sam-
ple. The skewness and kurtosis calculations help to evaluate the
distribution shape for each of the Rorschach variables. Variables
that exceed a skewness of 2 and/or kurtosis of 7 are considered to
be violations of normality (Curran, West, & Finch, 1996). Based
on these parameters, an examination of Table 2 shows that the
distributions of some variables exceed the boundaries of normal-
ity. In order to address this concern, any variables with a skewness
>2 and/or kurtosis >7 (Inner-Outer, Self-Other, Split) were log-
transformed to normalize their distribution. These log-transformed
variables were used in the rest of the reported analyses.
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 213
In order to evaluate the significance of differences among the
clinical and nonclinical groups, a series of ANOVAs were calcu-
lated for both demographic and Rorschach variables. A power
analysis using Cohen’s tables (1988) shows that a sample size of at
least 64 participants is needed to detect medium effects in a two-
group ANOVA. The sample size used for this analysis was 100 and
the results appear in Table 3. In certain cases, a positive correlation
between Rorschach variables and the number of responses (R) on
a given protocol may produce artificial differences between groups.
To avoid this complication, the effects of R were partialed out of
three variables (A2, Ideal, and ROD) that demonstrated a signifi-
cant relationship with the total number of responses. Results of the
ANOVAs for the Rorschach variables show that the BPD group
had significantly higher scores than the nonclinical sample with the
Table 3. Comparison of Non-clinical and Total BPD Sample (N = 100).
NC (n = 50) Total BPD (n = 50)
Variables X SD X SD p d#
Age 22.6 5.3 30.5* 9.3 <.0001 1.06
Education 14.8 1.1 13.9** 2.6 .01 .55
BDS
Inner-Outer^^ .31 .08 .59 .27 <.0001 1.43
Self-Other^^ .35 .16 .57 .26 <.0001 1.0
Aggression
A1 .06 .24 .94 .96 <.0001 1.3
A2^ 2.2 .98 3.3 1.8 <.0001 .75
LDS
Split^^ .36 0.0 .54 .25 <.0001 .96
Deval 3.1 3.1 8.6 5.8 <.0001 1.2
Ideal^ .09 .12 .16 .19 .02 .48
MOAS
MOAS-H 4.5* 1.8 6.0 1.3 <.0001 .95
MOAS-PATH .60* .64 2.7 1.9 <.0001 1.47
Dependency
ROD^ .16 .08 .19 .14 .17 .28
Note. ^Effects of R were partialed out of analyses. ^^Log-transformed scores. *n = 49. **n = 44.
#effect sizes reported according to Cohen (1977); small>.2; medium>.5; large >.8.
Baity et al.
214 Bulletin of the Menninger Clinic
exception of ROD. Effect size calculations were quite large (small
= >.2; medium = >.5; large = >.8; Cohen, 1977) for a majority of
the Rorschach variables (Inner-Outer, Self-Other, A1, Split, Deval,
MOAS-H, and MOAS-PATH) with an additional variable near-
ing a large effect size (A2 = .75). This highlights the strong pos-
sibility that true score differences exist. The nonclinical group was
significantly younger and more educated (although by only about
one year) than the clinical group. While the differences found be-
tween a college and a clinical sample is not groundbreaking, the
availability of nonclinical data for the psychodynamic scales used
in this study is virtually nonexistent. Readers should be cautious
about comparing results in Table 3 with outside data because the
means and SD of some variables are based on log-transformations.
Instead, an appendix (Appendix A) is provided with the untrans-
formed means and SD of each Rorschach variable for the clinical
groups reported on in this study.
Further exploration of the clinical and nonclinical groups was
done using a logistic regression analysIs to see how well the Ror-
schach scales could predict the total number of BPD criteria using
the full sample (N = 100). In other words, how well could clinical
protocols be separated from nonclinical ones using only the Ror-
schach scales? Because both Age and Education were significantly
different between the two groups, these were entered into Block
1 of the logistic regression. Each of the Rorschach variables was
then entered as predictors of the model in Block 2. The predictor
variables were entered in a forward manner until the inclusion of
variables no longer improved the significance of the model. Results
of the logistic regression (Table 4), in the order of appearance in
the model, indicated that both Age and Education were significant
predictors of total BPD criteria. Among the Rorschach variables,
the Inner-Outer boundary variable and Devaluation were also
nonredundant and independent predictors of group membership
above and beyond the variance accounted for by the demographic
variables. The Hosmer and Lemeshow (1989) Goodness-of-Fit
Test was not significant (X2 = 13.28; p = .10), indicating that the
regression model, with a Classification Accuracy of 87%, is a good
fit for the data. The Rorschach variables contributed above and
beyond Age and Education and suggest that a combination of both
greater boundary diffusion and more examples of Devaluation best
differentiated the clinical from the nonclinical group.
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 215
The next stage of the analysis included comparisons among those
BPD patients with and without a history of self-mutilation. Table 5
shows the means and standard deviations for the NC, non-self-mu-
tilating BPD (N-BPD; n = 35), and self-mutilating BPD (SM-BPD;
n = 15) groups, as well as comparisons between the NC group and
both of the BPD groups. When compared to the NC group, the
SM-BPD sample had higher means and larger effect sizes than the
N-BPD sample for a majority of the variables, suggesting that BPD
patients who self-mutilate are more disturbed than BPD patients
who do not engage in this behavior. The exception is the Inner-
Outer variable, which had a higher mean and larger effect size in
the N-BPD group. Overall, the results in Table 5 outline a clear
difference in pathology between each of the BPD groups and the
nonclinical sample.
The larger effect sizes for the SM-BPD group compared to the
N-BPD group were felt to warrant further investigation. Table 6
furthers the idea of a BPD pathology spectrum by comparing the
means and standard deviations of the Rorschach scales from the
N-BPD group and SM-BPD groups. Some differences did emerge
between the two clinical groups with moderately large effect sizes.
The SM-BPD had significantly higher means than the N-BPD on
ROD (p = .03), with nearly significant differences for A1 (p = .06)
and MOAS-H (p = .06). However, the smaller sample size for this
analysis (n = 50) results in a loss of power, thus requiring more cau-
tion when interpreting group differences.
A second logistic regression analysis was conducted determine
how well the Rorschach scales could differentiate BPD patients
Table 4. Logistic Regression Summary for DSM-IV BPD Diagnosis (N = 100).
Variable B SE Wald df p Exp(B) %
Block 1
Age .162 .045 13.123 1 >.001 1.176 76%
Education -.297 .138 4.617 1 .032 .743 76%
Block 2
Inner-Outer^^ 10.015 2.529 15.685 1 >.001 22365.264 82.6%
Deval .423 .123 11.793 1 .001 1.527 87%
Note. % = Increase in classification accuracy of the model. ^^Log-transformed scores. *N = 99.
Baity et al.
216 Bulletin of the Menninger Clinic
Table 5. Comparison of Non-clinical with N-BPD and SM-BPD Subsamples (N = 100).
NC (n = 50) N-BPD (n = 35) SM-BPD (n = 15) NC vs. N-BPD NC vs. SM-BPD
Variables X SD X SD X SD p d p d
Inner-Outer^^ .31 .08 .63 .26 .50 .28 <.0001 1.8 <.0001 1.3
Self-other^^ .35 .16 .56 .26 .59 .29 <.0001 1.0 .0001 1.2
A1 .06 .24 .77 .77 1.3 1.2 <.0001 1.3 <.0001 2.0
A2^ 2.2 .98 3.0 1.6 3.8 2.0 .005 .60 <.0001 1.2
Split^^ .36 0.0 .50 .23 .62 .29 .0001 .90 <.0001 1.9
Deval 3.1 3.1 8.2 5.6 9.7 6.3 <.0001 1.2 <.0001 1.6
Ideal^ .09 .12 .14 .16 .22 .25 .08 .4 .009 .8
MOAS-H 4.5* 1.8 5.7 1.5 6.5 .64 .001 .7 .0001 1.2
MOAS-PATH .60* .64 2.5 2.0 2.9 1.7 <.0001 1.4 <.0001 2.3
ROD^ .16 .08 .17 .11 .26 .16 .87 .10 .003 1.0
Note. NC = Nonclinical sample; N-BPD = Non-self-mutilating BPDs; SM-BPD = Self-mutilating BPDs. ^Effects of R were partialed out of analyses. ^^Log-transformed
scores. *N = 49.
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 217
who self-mutilate from those who do not. The presence or absence
of self-mutilation was entered as the criterion variable in these
analyses. Due to the reduced sample size, using all 10 original Ror-
schach scales would violate the minimally acceptable participant to
IV ratio of 10:1 in regression analyses. Therefore, the Rorschach
variable with the largest effect size from each category (BDS, Ag-
gression, LDS, MOAS, and ROD) in Table 6 served as the pre-
dictor variables. Those variables were Inner-Outer, A1, Splitting,
MOAS-H, and ROD. The remaining procedures for insertion and
interpretation of the second logistic regression analyses were the
same as described above, and the results are shown in Table 7.
Among the Rorschach scales entered into the analysis, ROD was
the only variable to make a significant, nonredundant contribu-
tion. Once again, the Hosmer and Lemeshow (1989) Goodness-
of-Fit Test was nonsignificant (X2 = 8.745; p = .364), suggesting
a good fit for the final model. The classification accuracy was a
modest 70% and was likely influenced by the small sample size of
the SM-BPD group. Despite the limited generalizability, ROD con-
sistently discriminated between the BPD groups on both ANOVA
and regression analyses, thus increasing confidence of true differ-
ences in larger samples. In addition to more obvious differences
between clinical and nonclinical samples, a distinct pattern of re-
Table 6. Comparison of N-BPD and SM-BPD Subsamples (n = 50)
N-BPD (n = 35) SM-BPD (n = 15)
Variables X SD X SD p d
Inner-Outer^^ .63 .26 .50 .28 .13 .48
Self-other^^ .56 .26 .59 .29 .78 .09
A1 .77 .77 1.3 1.2 .06 .62
A2^ 3.0 1.6 3.8 2.0 .13 .48
SPLIT^^ .50 .23 .62 .29 .13 .49
DEVAL 8.2 5.6 9.7 6.3 .39 .28
IDEAL^ .14 .16 .22 .25 .21 .40
MOAS-H 5.7 1.5 6.5 .64 .06 .62
MOAS-PATH 2.5 2.0 2.9 1.7 .51 .21
ROD^ .17 .11 .26 .16 .03 .71
Note. N-BPD = Non-self-mutilating BPDs; SM-BPD = Self-mutilating BPDs. ^Effects of R were
partialed out of analyses. ^^Log-transformed scores.
Baity et al.
218 Bulletin of the Menninger Clinic
sponding emerged within the two BPD groups where the pathology
of SM-BPD seems to indicate greater disruptions in interpersonal
functioning.
Discussion
The primary purpose of this study was to explore the continuum
of borderline pathology using variables shown in previous Ror-
schach assessment literature to have utility for making group dis-
tinctions. This study was the first attempt to replicate and extend
the findings of Fowler et al. (2000) using different samples and
statistical techniques. The current study is also distinctive in that it
is the first to report on nonclinical data for the psychoanalytically
derived Rorschach scales used in the current study. Large differ-
ences were expected and consistently found between the clinical
and nonclinical protocols in this study. Although not diagnostically
useful, nonclinical data help to establish baseline scores and serve
to anchor one end of the psychological developmental spectrum for
these variables. Given that nonclinical norms are usually some of
the first data collected for a new scale, it is a bit puzzling why these
data are just being reported for the first time. Anticipating that
clinical norms for the Rorschach scales used in this study may be of
use to future researchers, we provide a table (Appendix A) with the
unadjusted means and standard deviations for each of the scales
used in this study. Interestingly enough, no differences in ROD
scores emerged between the Total BPD and NC groups. The un-
adjusted mean ROD score (as shown in Appendix A) for the total
BPD sample in this study (X = 4.0) was low compared to the mean
ROD score from Fowler et al.'s inpatient data (2000; X = 5.9). The
difference in mean scores may be attributed to the use of both inpa-
tients and outpatients in the current sample. One study from which
data were used for the current analysis (Blais, Hilsenroth, Fowler,
et al., 1999) found a strong negative correlation between ROD and
Table 7. Logistic Regression Summary for Self-Mutilation (N = 50).
Variable B SE Wald df p R Exp(B) %
ROD^ 5.149 2.503 4.231 1 .04 .19 172.2682 70
Note.% = classification accuracy. ^Effects of R were partialed out of analyses.
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 219
total BPD criteria, suggesting that the sample of outpatients in that
study defended more rigidly against dependency needs.
The first logistic regression analysis in this study attempted to
predict clinical versus nonclinical group membership based on the
10 psychodynamic Rorschach scales selected a priori. The resulting
model (after the variances of Age and Education were accounted
for) predicted group membership with a total accuracy of 87%.
Increased frequency of Inner-Outer and Devaluation scores on the
Rorschach was more indicative of individuals diagnosed with BPD
than of college students. The finding that only two Rorschach vari-
ables entered into the regression model was a bit surprising given
the number and magnitude of differences observed in the ANOVAs
(Table 3). One possible explanation for this finding might be the
amount of variance accounted for by the demographic differences
between the groups. Future researchers are advised to match clinical
and nonclinical groups on certain variables such as age and educa-
tion. Inner-Outer boundary violations are proposed to occur when
a person’s perceptions of the internal and external worlds become
enmeshed (but not entirely inseparable) with one another. Such an
occurrence may lead one to confuse his or her own thoughts, feel-
ings, and wishes with those of other people. This has been dis-
cussed as highly prototypic of borderline pathology (Kernberg,
1975; Knight, 1953; Lerner et al., 1985). Devaluation has long
been associated with borderline character pathology history and is
typically thought of as a means to keep interpersonal distance from
the external world. Such uses of devaluation often occur when the
BPD patient realizes that idealized others are incapable of meeting
unrealistic needs. The subsequent rejection is often infused with
very strong, often aggressive, affect. In fact, recent studies with the
LDS have found that Devaluation is solely correlated with BPD
DSM-IV criteria when compared to other Axis II diagnoses, and
that the LDS Devaluation score predicts BPD DSM criteria above
the MMPI-2 BPD scale (Blais, Hilsenroth, Castlebury, Fowler, Ba-
ity, 1999; Blais et al., 2001).
The SM-BPD group overall appeared more disturbed in this
study when compared to the NC and N-BPD groups. This finding
replicates the results reported by Blais, Hilsenroth, Fowler, et al.
(1999) and Fowler et al. (2000) and reinforces the evidence that
BPD patients with a history of self-mutilation appear more psy-
chologically compromised on the Rorschach than BPD patients
Baity et al.
220 Bulletin of the Menninger Clinic
with no self-mutilative history. Additional ANOVAs between the
BPD groups revealed that ROD scores were significantly higher in
the SM-BPD group than in the N-BPD group. A1 and MOAS-H
scores were also very near the cutoff for significance, and these
scores might have shown greater differences with a larger sample
size in the SM-BPD group. One of the distinctive features of the
current study was the use of logistic regression analysis to predict
self-mutilation status of the two BPD groups. The resulting model
showed that ROD was the only nonredundant predictor of self-
mutilation and was able to classify the presence/absence of self-
mutilation 70% of the time.
Findings from the current study showed some distinct differ-
ences from Fowler et al. (2000) when comparing BPD patients with
and without self-mutilating behavior. With the exception of A1,
which approached statistical significance at p = .06 and had a me-
dium effect size, the remaining results did not seem to conform.
The only variable that was statistically significant between the two
BPD groups in the current study was ROD; however, this difference
was not found in Fowler et al. Despite the negative expectations in
relationships, the significantly higher ROD score for the SM-BPD
group suggests that these individuals may have a stronger pull to
depend on others than those with BPD who do not self-mutilate. It
would therefore be reasonable to assume that environmental trig-
gers for SM-BPD individuals occur primarily in the context of re-
lationships feeling either too distant or too close. As these triggers
build affective momentum, the likelihood of self-mutilation may
increase as a way to replace emotional pain with physical pain and/
or interrupt the dissociation commonly experienced with intense
emotional episodes (Fowler & Hilsenroth, 1999).
The data of Fowler et al. (2000) was based strictly on inpatients
who had been admitted to a long-term treatment facility for at least
6-months versus a combination of inpatients from a short-term
psychiatric facility and outpatients used for the current study. Pa-
tients from the Fowler et al. study are likely more pathological than
the sample used for this study so that even BPD patients without
a history of self-harming behavior struggle with issues of depen-
dency (higher ROD scores). At the other end of the continuum is
the significant negative relationship found between ROD and total
BPD criteria in outpatients, suggesting a more tightly regimented
interpersonal world (Blais, Hilsenroth, Fowler, et al., 1999) in pa-
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 221
tients taken from this sample for the current study. The combined
sample used in this study likely falls between these two points and
may have inadvertently diluted the data. As borderline pathology
increases in severity of impairment, it may be that the vigilantly
guarded feelings of neediness seen with low ROD scores begin to
transform into strong feelings of avoiding abandonment (e.g., high
ROD). This finding further supports the idea that borderline pa-
thology has a spectrum of dysfunction, even among patients who
self-mutilate.
The predictions that the Rorschach would be able to establish a
range of pathology, distinguish clinical from nonclinical protocols,
and identify BPD patients’ self-mutilative status were all supported
in the current study. The two greatest limitations on the results
reported herein seem to be the small group sizes and the heteroge-
neity of the BPD group. The lower power of the analyses compar-
ing the two self-mutilating groups (N-BPD and SM-BPD) would
recommend that caution be used when interpreting results, espe-
cially those with small to low- moderate effect sizes. Obviously,
replications with larger samples are necessary to further examine
the work of Fowler et al. (2000). Given the wide range of pathol-
ogy assumed under both the DSM-IV BPD and Kernberg’s BPO, it
is unrealistic to expect a single pattern or collections of test signs to
identify (classify) or describe all members of such a heterogeneous
group of patients with complete accuracy. Despite these confounds,
distinct differences were found using the Rorschach that have also
been reported in previous theoretical and empirical writings and
seems to point to specific areas of functioning that are impaired by
BPD. Systematically reviewing performance-based data to deter-
mine a patient’s level of functioning across these psychological do-
mains and then organizing these observations into a coherent psy-
chological picture should enhance the clinician’s ability to identify
borderline psychopathology and describe aspects of psychological
functioning that greatly affect nontest behavior. Approaching Ror-
schach data in this manner will also allow clinicians to estimate the
severity of a patient’s condition as well as make meaningful predic-
tions regarding treatment.
Baity et al.
222 Bulletin of the Menninger Clinic
References
American Psychiatric Association. (1980). Diagnostic and statistical manual of
mental disorders (3rd ed.). Washington, DC: Author.
American Psychiatric Association. (1994). Diagnostic and statistical manual of
mental disorders (4th ed.). Washington, DC: Author.
Baity, M.R., & Hilsenroth, M.J. (1999). Rorschach aggression variables: A study
of reliability and validity. Journal of Personality Assessment, 72, 93-110.
Berg, J.L., Packer, A., & Nunno, V.J. (1993). A Rorschach analysis: Parallel
disturbance in thought and self object representation. Journal of Personality
Assessment, 61, 311-323.
Blais, M.A., & Baity, M.R. (2008). The projective assessment of personality
structure and pathology. In G.J. Boyle, G. Matthews, & D.H. Saklofske
(Eds.), The sage handbook of personality theory and assessment: Vol. 2. Per-
sonality measurement and testing (pp. 566–586). Thousand Oaks, CA: Sage
Publications.
Appendix A. Unadjusted Means and SD for NC and BPD groups (N = 100)
NC
(n = 50) Total BPD
(n = 50) N-BPD
(n = 35) SM-BPD
(n = 15)
Variables X SD X SD X SD X SD
BDS
Inner-Outer .12 .33 1.7 1.9 1.9 1.9 1.3 2.0
Self-Other .20 .45 .98 1.24 .86 .88 1.3 1.8
Aggression
A1 .06 .24 .94 .96 .77 .77 1.3 1.2
A2 4.4 2.0 6.5 3.6 6.0 3.3 7.7 4.0
LDS
Split .00 0.0 .46 .89 .34 .76 .73 1.1
Deval 3.1 3.1 8.6 5.8 8.2 5.6 9.7 6.3
Ideal 1.9 2.5 3.6 4.0 3.2 3.7 4.5 4.8
MOAS
MOAS-H 4.5 1.8 6.0 1.3 5.7 1.5 6.5 .64
MOAS-PATH .60 .64 2.7 1.9 2.5 2.0 2.9 1.7
Dependency
ROD 3.5 1.6 4.0 2.9 3.5 2.8 5.2 2.7
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 223
Blais, M.A., & Bistis, K. (2004). The projective assessment of borderline psycho-
pathology. In M.J. Hilsenroth & D.L. Segal (Eds.), Comprehensive handbook
of psychological assessment (Vol. 2, pp. 485-499). Hoboken, NJ: Publisher.
Blais, M.A., Hilsenroth, M.J., & Castlebury, F.D. (1997). Content validity of the
DSM-IV borderline and narcissistic personality disorder criteria sets. Compre-
hensive Psychiatry, 38, 31-37.
Blais, M.A., Hilsenroth, M.J., Castlebury, F., Fowler, J.C., & Baity, M.R. (2001).
Predicting DSM-IV cluster b personality disorder criteria from MMPI-2 and
Rorschach data: A test of incremental validity. Journal of Personality Assess-
ment, 76, 150-168.
Blais, M.A., Hilsenroth, M.J., & Fowler, J.C. (1999). Diagnostic efficiency and
hierarchical functioning of the DSM-IV borderline personality disorder crite-
ria. Journal of Nervous and Mental Disease, 187, 167-173.
Blais, M.A., Hilsenroth M.J., Fowler, J.C., & Conboy, C.A. (1999). A Rorschach
exploration of the DSM-IV borderline personality disorder. Journal of Clinical
Psychology, 55, 1-10.
Blatt, S.J., & Berman, W. (1984). A methodology for the use of the Rorschach in
clinical research. Journal of Personality Assessment, 48, 226-239.
Blatt, S., & Ritzler, B. (1974). Thought disorder and boundary disturbance in
psychosis. Journal of Consulting and Clinical Psychology, 42, 370-381.
Bornstein, R.F. (1996). Construct validity of the Rorschach oral dependency
scale: 1967–1995. Psychological Assessment, 8, 200-205.
Bornstein, R.F., Hilsenroth, M.J., Padawer, J.R., & Fowler, J.C. (2000). Interper-
sonal dependency and personality pathology: Variations in Rorschach Oral
Dependency scores across axis II diagnoses. Journal of Personality Assess-
ment, 75, 478-491.
Cohen, J. (1977). Statistical power analysis for the behavioral sciences (2nd ed.).
New York: Academic Press.
Cohen, J. (1988). A power primer. Psychological Bulletin, 112, 155-159.
Curran, P., West, S., & Finch, J. (1996). The robustness of test statistics to non-
normality and specification error in confirmatory factor analysis. Psychologi-
cal Methods, 1, 16-29.
Doctors, S. (1981). The symptom of delicate self-cutting in adolescent females: A
developmental view. Adolescent Psychiatry, 9, 443-460.
Exner, J. (1993). The Rorschach: A comprehensive system. Vol. 1. Basic founda-
tions (3rd ed.). New York: Wiley.
Fleiss, J. (1981). Statistical methods for rates and proportions (2nd ed.). New
York: Wiley.
Fowler, J.C., & Hilsenroth, M.J. (1999). Some reflections on self-mutilation.
Psychoanalytic Review, 86, 721-731.
Fowler, J.C, Hilsenroth, M.J., & Handler, L. (1995). Early memories: An explo-
ration of theoretically derived queries and their clinical utility. Bulletin of the
Menninger Clinic, 59, 79-98.
Fowler, J.C., Hilsenroth, M.J., & Handler, L. (1996). A multimethod approach
to assessing dependency: The early memory dependency probe. Journal of
Personality Assessment, 67, 399-413.
Baity et al.
224 Bulletin of the Menninger Clinic
Fowler, J.C, Hilsenroth, M.J., & Handler, L. (1998). Assessing transitional phe-
nomena with the transitional object memory probe. Bulletin of the Menninger
Clinic, 62, 455-474.
Fowler, J.C., Hilsenroth, M.J., & Nolan, E. (2000). Exploring the inner world of
self-mutilating borderline patients: A Rorschach investigation. Bulletin of the
Menninger Clinic, 64, 365-385.
Gunderson, J., G. (2001). Borderline personality disorder: A clinical guide. Wash-
ington, DC: American Psychiatric Publishing, Inc.
Hilsenroth, M.J., Hibbard, S.R., Nash, M.R., & Handler, L. (1993). A Ror-
schach study on narcissism, defense, and aggression in borderline, narcissistic,
and Cluster C personality disorders. Journal of Personality Assessment, 60,
346-361.
Holt, R. (1977). A method for assessing primary process manifestations and their
control in Rorschach responses. In M. Ricker-Ovsinkina (Ed.), Rorschach
psychology (2nd ed., pp. 263-315). Huntington, NY: Krieger.
Hosmer, D.W., & Lemeshow, S. (1989). Applied logistic regression. New York:
Wiley.
Kernberg, O. (1970). A psychoanalytic classification of character pathology. Jour-
nal of the American Psychoanalytic Association, 18, 800-822.
Kernberg, O. (1975). Borderline conditions and pathological narcissism. New
York: Aronson.
Kernberg, O. (1984). Severe personality disorders: Psychotherapeutic strategies.
New Haven, CT: Yale University Press.
Kessel, J., & Zimmerman, M. (1993). Reporting errors in studies of diagnostic
performance of self-administered questionnaires: Extent of the problem, rec-
ommendations for standardized presentation of results, and implications for
the peer review process. Psychological Assessment, 53, 395-399.
Knight, R.P. (1953). Borderline states. Bulletin of the Menninger Clinic, 17, 1-12.
Lerner, H.D., Sugarman, A., & Barbour, C.G. (1985). Patterns of ego boundary
disturbance in neurotic, borderline, and schizophrenic patients. Psychoana-
lytic Psychology, 2, 47-66.
Lerner, P. (1991). Psychoanalytic theory and the Rorschach. HillSDale, NJ:
Analytic Press.
Lerner, P., & Lerner, H. (1980). Rorschach assessment of primitive defense in
borderline personality structure. In J. Kwawer, H. Lerner, P. Lerner, & A. Sug-
arman (Eds.), Borderline phenomena and the Rorschach test (pp. 257-274).
New York: International Universities Press.
Masling, J., Rabie, L., & Blondheim, S. (1967). Obesity, level of aspiration, and
Rorschach and TAT measures of oral dependence. Journal of Consulting
Psychology, 31, 233-239.
McWilliams, N. (2004). Psychoanalytic diagnosis. New York, NY: The Guilford
Press.
Quinlan, D., & Harrow, M. (1974). Boundary disturbance in schizophrenia.
Journal of Abnormal Psychology, 83, 533-541.
Rapaport, D., Gill, M., & Schafer, R. (1945). Diagnostic psychological testing.
Chicago: Yearbook Publishers.
Rorschach and Self-Mutilation
Vol. 73, No. 3 (Summer 2009) 225
Rapaport, D., Gill, M., & Schafer, R. (1968). Diagnostic psychological testing
(R. Holt, Ed.; rev. ed.). New York: International Universities Press.
Stolorow, R.D., & Lachmann, F.M. (1980). Psychoanalysis of developmental ar-
rests: Theory and treatment. New York: International Universities Press.
Urist, J. (1977). The Rorschach test and the assessment of object relations. Jour-
nal of Personality Assessment, 41, 3-9.
Weiner, I. (1991). Editor’s note: Interscorer agreement in Rorschach research.
Journal of Personality Assessment, 56, 1.
