Revising and Assessing Axis II,

Part II: Toward an Empirically Based and Clinically Useful

Classification of Personality Disorders

Drew Westen, Ph.D., and Jonathan Shedler, Ph.D.

Objective: The DSM-IV classification of personality disorders has not proven satisfying
to either researchers or clinicians. Incremental changes to categories and criteria using
structured interviews may no longer be useful in attempting to refine axis II. An alternative
approach that quantifies clinical observation may prove useful in developing a clinically
rich, useful, empirically grounded classification of personality pathology. Method: A total of
496 experienced psychiatrists and psychologists used the Shedler-Westen Assessment
Procedure-200 (SWAP-200) to describe current patients diagnosed with axis II personality
disorders. The SWAP-200 is an assessment tool that allows clinicians to provide detailed,
clinically rich descriptions of patients in a systematic and quantifiable form. A statistical
technique, Q-analysis, was used to identify naturally occurring groupings of patients with
personality disorders, based on shared psychological features. The resulting groupings
represent an empirically derived personality disorder taxonomy. Results: The analysis
found 11 naturally occurring diagnostic categories, some of which resembled current axis
II categories and some of which did not. The findings suggest that axis II falls short in its
attempt to “carve nature at the joints”: In some cases it puts patients who are psychologi-
cally dissimilar in the same diagnostic category, and in others it makes diagnostic distinc-
tions where none likely exist. It also fails to recognize a large category of patients best char-
acterized as having a dysphoric personality constellation. The empirically derived
classification system appears to be more faithful to the clinical data and to avoid many
problems inherent in the current axis II taxonomy. Conclusions: The approach presented
here may be helpful in refining the existing taxonomy of personality disorders and moving
toward a system of classification that lies on a firmer clinical and empirical foundation. In
addition, it can help to bridge the gap that often exists between research and clinical ap-
proaches to personality pathology.
(Am J Psychiatry 1999; 156:273–285)

A xis II of DSM-IV represents a hybrid of clinical

and research observations. The diagnostic categories
pirically based changes in axis II have clearly improved
the personality disorder taxonomy. However, they
have their origins in clinical observation and theory, have truly satisfied neither researchers nor clinicians,
and the categories and criteria have been refined over including members of the DSM-IV task force itself,
the years through empirical research. The gradual, em- some of whom have called for the elimination of the
current categorical system in favor of a dimensional
Received Jan. 14, 1998; revision received June 22, 1998; system (see reference 1).
accepted Aug. 25, 1998. From the Department of Psychiatry, Har- The methods currently used to revise axis II have a
vard Medical School, Boston; and The Cambridge Hospital/Cam-
number of limitations. For our purposes, the most im-
bridge Health Alliance. Address reprint requests to Dr. Westen,
Department of Psychiatry, The Cambridge Hospital, 1493 Cam- portant are the following (see also part I of this two-
bridge St., Cambridge, MA 02139; [email protected] (e-mail). part series).
The authors acknowledge the assistance of the over 950 clini- 1. Current personality disorder instruments have
cians who helped to refine the SWAP-200 assessment instrument, significant empirical and conceptual limitations. For
including the 797 who participated in the present study. They also
thank several research assistants who helped in the collection of example, they have marginal validity and poor retest
the data, particularly Michelle Levine, Alan Reyes, Lisa Goldstein, reliability at intervals greater than 6 weeks. An addi-
and Elizabeth Schafer. tional problem is that these instruments do not mirror

Am J Psychiatry 156:2, February 1999 273

REVISING AND ASSESSING AXIS II, PART II

the assessment procedures used in clinical practice. Cli- slightly more differentiated model, arguing that case
nicians typically assess personality by listening to pa- formulations should address three broad questions: 1)
tients’ narrative accounts of their experiences, noting What does the person wish for and fear, and to what
their behavior in the consulting room, and then mak- extent are these wishes and fears conflicting or uncon-
ing inferences about personality processes. In contrast, scious? 2) What psychological resources—cognitive,
current instruments rely on direct questions and expect affective, and behavioral—can the person draw upon
patients to report on their own personalities. It is to meet internal and external demands? 3) How does
highly unlikely that most patients with personality dis- the person perceive and experience self and others,
orders can do so adequately (see part I). and how able is he or she to sustain meaningful and
The reliance on such instruments to refine axis II pleasurable relationships?
criteria has led to an inversion of the normal proce- If the concept “personality” subsumes such domains
dures for selecting diagnostic criteria. Instead of iden- of functioning, then current axis II criteria for many
tifying the best diagnostic criteria and then finding disorders do not provide even a minimal outline for de-
ways to operationalize them, axis II committees have scribing a personality style. Consider, for example,
tended to exclude criteria that cannot be assessed by paranoid personality disorder, which is currently de-
direct questions (for several examples, see reference fined by the following criteria: fears of deceit or exploi-
2). Thus, we may be limiting the clinical applicability tation, fears of betrayal, fears that others will use in-
of DSM by linking its refinement so closely to a par- formation against them, fears that people have hidden
ticular method of assessment. hostile meanings in their communications, fears that
2. Current instruments are too wedded to the exist- people are attacking them, fears of infidelity, and a ten-
ing taxonomy. The questions included in current as- dency to hold grudges against people perceived as hav-
sessment instruments are derived from existing diag- ing done such things. These criteria are essentially
nostic criteria and therefore are of limited value for seven indices of a single trait, chronic mistrust. One of
developing new or better criteria. Most efforts at refin- them (fear of infidelity) is not empirically related to the
ing axis II criteria examine relations between potential disorder but was maintained as a criterion because it
new criteria and existing axis II criteria. The problem seemed to express one more type of malevolent con-
with this approach is that it assumes we already have cern (8).
the categories and general constellations of symptoms Knowing that a person tends to be distrustful in mul-
right, since potential criteria are excluded if they do tiple ways, however, says little about his characteristic
not correlate highly with existing criteria (or if they ways of thinking (How disordered can his thought be-
correlate too highly with criteria for other disorders). come? Is it disordered in noninterpersonal realms as
Since neither the personality disorder categories nor well?), the feelings he typically experiences (Is he sad?
the constellations of diagnostic criteria were estab- Is he shame-prone? Is he anxious?), the ways he deals
lished empirically, and since they typically do not with those feelings (Does he attempt to manage them
closely match the results of cluster and factor analyses by seeking information, by using substances, by turn-
(3, 4), we may at times be refining item sets to fit cat- ing to a confidante who remains outside his malevolent
egories and criteria that exist by convention. With use thought system, by developing grandiose ideas about
of current methods for refining axis II, there is no his place in the world, by projecting his feelings onto
way to solve this problem. Any alternative diagnostic others?), what he wishes for in life, what skills he has,
category that better distinguished groups of patients how he sees himself, how he spends his time, and so
currently classified into existing categories could not on. Such questions are crucial clinically because they
be discovered or implemented because its criteria provide insight into the possible functions of the pa-
would necessarily overlap with current criteria from tient’s symptoms, as well as the psychological strengths
other disorders, which may themselves be somewhat and weaknesses that bear on the person’s adaptation to
arbitrary. life.
3. The current diagnostic categories do not encom- 4. Axis II criteria are becoming increasingly narrow.
pass the domains of functioning relevant to personal- Although not the intent of the axis II work groups, the
ity. The architects of the DSM system have attempted methods used to revise and refine axis II inherently
to avoid diagnostic criteria that are tied too closely to lead to ever-narrower diagnostic criteria, which cap-
any particular theoretical orientation. This is clearly ture less and less of the richness and complexity of the
sensible, since the diagnostic manual must be useable clinical data. The reasons for this have not, we believe,
by clinicians of all theoretical orientations. However, been adequately appreciated.
it has left axis II committees without guidance regard- Axis II work groups have labored under the con-
ing the domains of functioning relevant to the concept straint of trying to 1) maximize the internal consis-
of “personality.” Personality psychologists continue to tency of the diagnostic criteria for each disorder (i.e.,
debate the precise definition of personality, but most the correlations between them) and 2) reduce correla-
agree it refers to the interaction of enduring patterns of tions with criteria for other disorders, while 3) limiting
1) cognition, 2) emotion, 3) motivation, and 4) behav- themselves to only seven to 10 diagnostic criteria per
ior expressed under particular circumstances (see refer- disorder. In practice, this means that personality char-
ences 5 and 6). Elsewhere Westen (6, 7) has offered a acteristics relevant to multiple disorders must be ex-

274 Am J Psychiatry 156:2, February 1999

 DREW WESTEN AND JONATHAN SHEDLER

cluded from all diagnostic categories except one, to METHOD

avoid problems of comorbidity. For example, lack of
empathy was excluded as a criterion for antisocial per- Subjects and procedures were the same as those described in part
I. A total of 797 experienced psychologists and psychiatrists, drawn
sonality disorder to reduce comorbidity with narcissis- from a random national sample, used the SWAP-200 to provide de-
tic personality disorder, even though research has tailed descriptions of actual and hypothetical patients. The study re-
shown it is one of the most characteristic features of ported here is based on SWAP-200 descriptions of 496 actual pa-
tients, diagnosed by their clinicians as meeting axis II criteria for a
antisocial patients (9). personality disorder diagnosis. (Hypothetical patients and healthy,
When diagnostic criteria are revised to increase in- high-functioning patients were excluded from the group.)
To identify naturally occurring groupings among the personality
ternal consistency, the result is that the criteria become disorder patients, we used a procedure known as “Q” factor analy-
narrower in scope. This is inevitable, because it is psy- sis, or simply Q-analysis. Q-analysis was originally used by biolo-
chometrically impossible for seven to 10 items (crite- gists conducting taxonomic research, to help classify species. The
procedure identifies groups of patients who are similar to one an-
ria) to encompass a complex psychological construct other and dissimilar to patients in other groups. The technique has
such as a personality disorder and also have high inter- been used successfully in studies of normal personality (10–16) but
nal consistency. That is why personality researchers do not in studies of personality disorders.
not typically design personality tests with only seven to Q-analysis can be understood by comparison with conventional
factor analysis, which is a common statistical technique in psycholog-
10 items. Efforts to maximize the internal consistency ical research. Factor analysis is used when a data set contains many
of such a small number of diagnostic criteria inherently variables, and these variables appear to be redundant measures of a
lead to criteria that are redundant indices of a single few underlying dimensions (factors). The technique identifies groups
of variables that are highly similar to one another (i.e., highly corre-
trait, not descriptors of a personality configuration. lated) but unrelated to variables in other groups. A researcher can
Consider the following: If a personality disorder de- then examine the variables in each group to draw conclusions about
scription should include, at minimum, criteria relevant the underlying factor that they measure. (For example, if a group con-
tains variables such as “is often sad,” “has little interest in activities,”
to a person’s characteristic patterns of 1) thought, 2) af- “cries easily,” and “has suicidal thoughts,” the researcher may con-
fectivity, 3) motivation, and 4) behavior, then an eight- clude that they measure the underlying factor of depression.)
item test will contain, on average, only two items per Q-analysis (as used in this study) is computationally the same pro-
cedure as conventional factor analysis, except that it creates group-
domain of functioning. Psychometrically, 10 eight-item ings of similar people, not variables. Thus, Q-analysis identifies
tests (criteria sets) that each include four two-item groups of patients who share important psychological features that
“subscales” can never achieve acceptable internal con- distinguish them from patients in other groups. The groups, called
Q-factors, represent empirically derived diagnostic categories that
sistency and discriminant validity; no amount of tink- may represent a potential alternative to axis II. (In a typical data file,
ering with the item sets can overcome what is essen- columns represent variables and rows represent people. Factor anal-
tially a mathematical impossibility. ysis identifies columns of data that are similar to one another,
whereas Q-analysis identifies rows of data that are similar. The com-
putational procedure is identical, and is accomplished simply by in-
verting the data matrix [i.e., exchanging rows and columns] before
OVERVIEW OF THE PRESENT STUDY performing calculations.)
The Q-analysis we will present gauges the similarity (or dissimi-
larity) of patients by the correlation between their SWAP-200 de-
The question, then, is how to develop a classification scriptions (see part I). Note that the Q-analysis makes use of all 200
system for personality disorders that is 1) clinically items in the SWAP-200 to gauge the similarity of patients and thus
takes account of the configuration of personality characteristics
useful and faithful to the data of clinical observation across a broad range of items. The items assess multiple domains of
(since ultimately the diagnostic manual must apply to functioning, encompassing characteristic patterns of thought, feel-
patients in clinical practice), and 2) based on empirical ing, motivation, and behavior.
findings so it reflects as accurately as possible the cate-
gories of personality dysfunction that occur “in na-
RESULTS
ture.” This article represents one such effort. We re-
port findings based on a large group of personality Q-Analysis Procedure
disorder patients in treatment with an experienced psy-
chiatrist or psychologist drawn from a random na- The Q-analysis followed commonly accepted factor
tional sample. The clinicians described their patients analytic procedures; readers familiar with factor anal-
through use of the Shedler-Westen Assessment Proce- ysis will recognize the approach. To determine the
dure-200 (SWAP-200), an assessment tool that allows number of Q-factors to extract, we performed an ini-
tial principle components analysis and retained Q-fac-
clinicians to provide detailed and clinically rich psy-
tors (principal components) with eigenvalues of 1 or
chological descriptions of patients, in a systematic and
higher (Kaiser’s criteria). The procedure resulted in 14
quantifiable form (see part I). Our aim was to discover Q-factors, which collectively accounted for 57.2% of
whether this information could be used to identify clin- the variance in the data set. These Q-factors were then
ically and theoretically meaningful categories of per- subjected to varimax rotation (i.e., orthogonal rota-
sonality disorder patients, without assuming the cur- tion, designed to create independent or uncorrelated
rent axis II taxonomy a priori. Q-factors). The first seven of the rotated Q-factors

Am J Psychiatry 156:2, February 1999 275

REVISING AND ASSESSING AXIS II, PART II

TABLE 1. SWAP-200a Items That Best Describe Patients in Dys- equivalent to factor scores in conventional factor anal-
phoric Personality Disorder Diagnostic Category (Q-factor 1) ysis, except that they apply to items, not subjects.) The
Factor items are arranged in descending order of importance,
Itemb Score from highest to lowest.
Tends to feel he/she is inadequate, inferior, As the items in tables 1 through 7 make clear, there
or a failure. 3.62886
Tends to feel unhappy, depressed, or despondent. 3.11144
is little doubt about the interpretation of the Q-factors
Tends to feel ashamed or embarrassed. 2.75582 (diagnostic categories) or the appropriate names for
Tends to blame self or feel responsible for bad things them. The results suggest that the Q-analysis identi-
that happen. 2.70904 fied clinically and theoretically meaningful diagnostic
Tends to feel guilty. 2.67309 categories.
Tends to fear he/she will be rejected or abandoned by
those who are emotionally significant. 2.65899 Several aspects of the Q-factors are worthy of note.
Tends to feel helpless, powerless, or at the mercy of First, many of the categories clearly resemble current
forces outside his/her control. 2.52297 axis II diagnostic categories. However, the Q-factors
Tends to be overly needy or dependent; requires ex- have an important advantage over axis II categories,
cessive reassurance or approval. 2.30493
Tends to be ingratiating or submissive (e.g., may con-
namely, they reflect the empirical solution that maxi-
sent to things he/she does not agree with or does not mizes their distinctiveness and minimizes comorbidity.
want to do, in the hope of getting support (This result was ensured by the statistical technique
or approval). 2.12255 used in the Q-analysis, specifically varimax rotation of
Tends to be passive and unassertive. 2.11851 the Q-factors.) Thus, the typical personality disorder
Tends to be self-critical; sets unrealistically high stan-
dards for self and is intolerant of own human defects. 2.02035 patient will have one personality disorder diagnosis in
Tends to feel like an outcast or outsider; feels as if he/ the empirically derived typology, not the multiple diag-
she does not truly belong. 1.94180 noses common with axis II.
Tends to be anxious. 1.90766 Second, the largest number of patients—over 20%
Tends to feel listless, fatigued, or lacking in energy. 1.78595
Tends to feel empty or bored. 1.76592 of our group—were classified as belonging in the first
Appears to want to “punish” self; creates situations Q-factor, which is not in DSM-IV, and which we la-
that lead to unhappiness or actively avoids opportu- beled dysphoric personality disorder. Patients in this
nities for pleasure and gratification. 1.70725 Q-factor feel distressed in multiple ways and experi-
Appears to find little or no pleasure, satisfaction, or en-
joyment in life’s activities. 1.70711
ence feelings of inadequacy, shame, guilt, depression,
Tends to be insufficiently concerned with meeting own anxiety, and fear of rejection or abandonment. The
needs; appears not to feel entitled to get or ask for category included many patients diagnosed by their
things he/she deserves. 1.70099 treating clinician as having depressive, dependent,
Is unable to soothe or comfort self when distressed; re- avoidant, self-defeating, or borderline personality dis-
quires involvement of another person to help regu-
late affect. 1.59890 order. The finding suggests that many patients cur-
Lacks a stable image of who he/she is or would like to rently given these diagnoses belong to a single diag-
become (e.g., attitudes, values, goals, and feelings nostic group that is characterized by a dysphoric or
about self may be unstable and changing). 1.54911 depressive character structure. Patients in this dys-
Tends to feel life has no meaning. 1.48654
Tends to avoid social situations because of fear of em-
phoric category differ in the activating conditions for
barrassment or humiliation. 1.42795 their dysphoria (e.g., some become distressed when
Has difficulty acknowledging or expressing anger. 1.33819 forced to interact with other people, whereas others
a Shedler-Westen Assessment Procedure—200. become distressed when they feel alone) and in the
b Items listed in descending order of diagnostic import. ways they attempt to regulate it (e.g., by avoiding peo-
ple and situations, desperately clinging to others, an-
grily attacking others who frustrate them), but they
were theoretically coherent and readily interpretable
share the core characteristics of dysphoric affect and
and accounted for 48.4% of the variance in the data
self-condemnation.
set. Thus, we retained these seven Q-factors. Most of
Third, the Q-analysis treated three sets of disorders
the Q-factors included 40 or more patients with factor
loadings of 0.50 or higher. The seventh factor was rep- differently than axis II, in ways that appear better to
resented by 10 patients, with factor loadings ranging “carve nature at the joints.” A single schizoid Q-factor
from 0.40 to 0.59. Similar Q-factors emerged when we emerged that included many patients currently diag-
nosed as schizoid and schizotypal, as well as a subset
rotated different numbers of factors, although the so-
of patients currently diagnosed as avoidant. The dis-
lution described here yielded the most clinically coher-
ent findings. tinction between these three disorders has been a mat-
ter of controversy since the introduction of axis II, and
Empirically Derived Diagnostic Categories our data, like those of others (see reference 2), do not
support the current taxonomy. Rather, they suggest
Tables 1–7 list the SWAP-200 items that best de- that these categories do not describe distinct personal-
scribe the patients in each of the seven Q-factors or di- ity styles. A second divergence from axis II was that
agnostic categories. The second column shows the fac- patients currently diagnosed as borderline tended to
tor score for each item, which indicates its centrality or fall into either the dysphoric or histrionic Q-factors.
importance in defining the Q-factor. (The scores are This finding directly replicated the results of a pilot

276 Am J Psychiatry 156:2, February 1999

 DREW WESTEN AND JONATHAN SHEDLER

TABLE 2. SWAP-200a Items That Best Describe Patients in Anti- TABLE 3. SWAP-200a Items That Best Describe Patients in
social-Psychopathic Disorder Diagnostic Category (Q-factor 2) Schizoid Personality Disorder Diagnostic Category (Q-factor 3)
Factor Factor
Itemb Score Itemb Score
Tends to be deceitful; tends to lie or mislead. 3.22828 Lacks close friendships and relationships. 3.82667
Takes advantage of others; is out for number one; has Appears to have a limited or constricted range
minimal investment in moral values. 2.89580 of emotions. 3.31256
Appears to experience no remorse for harm or injury Lacks social skills; tends to be socially awkward or in-
caused to others. 2.53806 appropriate. 3.31193
Tends to be angry or hostile (whether consciously or Appearance or manner seems odd or peculiar (e.g.,
unconsciously). 2.46672 grooming, hygiene, posture, eye contact, speech
Tends to act impulsively, without regard rhythms, etc. seem somehow strange or “off”). 2.99736
for consequences. 2.41031 Tends to be shy or reserved in social situations. 2.55474
Tries to manipulate others’ emotions to get what he/ Tends to be inhibited or constricted; has difficulty al-
she wants. 2.34767 lowing self to acknowledge or express wishes
Tends to be unreliable and irresponsible (e.g., may fail and impulses. 2.32617
to meet work obligations or honor financial Has difficulty making sense of other people’s behavior;
commitments). 2.27594 often misunderstands, misinterprets, or is confused
Tends to engage in unlawful or criminal behavior. 2.23755 by others’ actions and reactions. 2.21872
Has little empathy; seems unable to understand or re- Appears unable to describe important others in a way
spond to others’ needs and feelings unless they that conveys a sense of who they are as people; de-
coincide with his/her own. 2.19878 scriptions of others come across as two-dimensional
Tends to be unconcerned with the consequences of and lacking in richness. 2.21060
his/her actions; appears to feel immune Has little psychological insight into own motives, be-
or invulnerable. 2.11653 havior, etc.; is unable to consider alternative interpre-
Tends to show reckless disregard for the rights, prop- tations of his/her experiences. 2.11123
erty, or safety of others. 2.09892 Tends to think in concrete terms and interpret things in
Tends to abuse alcohol. 2.00997 overly literal ways; has limited ability to appreciate
Tends to blame others for own failures or shortcom- metaphor, analogy, or nuance. 2.06673
ings; tends to believe his/her problems are caused by Appears to have little need for human company or con-
external factors. 1.99449 tact; is genuinely indifferent to the presence
Tends to get into power struggles. 1.95448 of others. 2.01159
Appears to gain pleasure or satisfaction by being sa- Perception of reality can become grossly impaired un-
distic or aggressive toward others (whether der stress (e.g., may become delusional). 1.91400
consciously or unconsciously). 1.82434 Tends to avoid social situations because of fear of em-
Has little psychological insight into own motives, be- barrassment or humiliation. 1.89556
havior, etc.; is unable to consider alternative interpre- Reasoning processes or perceptual experiences seem
tations of his/her experiences. 1.82308 odd and idiosyncratic (e.g., may make seemingly ar-
Repeatedly convinces others of his/her commitment to bitrary inferences; may see hidden messages or
change but then reverts to previous maladaptive be- special meanings in ordinary events). 1.86896
havior; tends to convince others that “this time is re- Tends to elicit boredom in others (e.g., may talk inces-
ally different.” 1.75198 santly, without feeling, or about inconsequential
a Shedler-Westen Assessment Procedure—200. matters). 1.74713
b Items listed in descending order of diagnostic import. Has difficulty acknowledging or expressing anger. 1.72157
Has difficulty allowing self to experience strong plea-
surable emotions (e.g., excitement, joy, pride). 1.71361
study of cluster B disorders (N=153), which used an Tends to be passive and unassertive. 1.71302
Tends to feel like an outcast or outsider; feels as if he/
earlier version of the SWAP (17). A third divergence she does not truly belong. 1.67419
from axis II was that a large percentage of patients a Shedler-Westen Assessment Procedure—200.
currently diagnosed as having obsessive-compulsive b Items listed in descending order of diagnostic import.

personality disorder appear to be substantially less dis-

turbed than the current axis II conceptualization. These
Subclassifying Within the Dysphoric Category
patients resemble Shapiro’s description of the obsessive
“neurotic style” (18) more than they resemble the obses- Because the first Q-factor contained so many pa-
sive-compulsive personality disorder of axis II. They are
tients, we conducted a second Q-analysis to identify
emotionally constricted, prone to intellectualization,
subgroups of patients within the dysphoric Q-factor.
and overly concerned with rules, but they are not partic-
The Q-analysis paralleled the Q-analysis procedure de-
ularly dysfunctional and they are conscientious and pro-
ductive to a fault. (We did identify an eighth Q-factor, scribed earlier. We performed a principal components
which we have not presented here because it contained analysis and retained eight Q-factors (principal com-
only seven patients of the 496 in the group. Patients in ponents) with eigenvalues greater than 1. These factors
this factor did resemble the DSM-IV description of ob- were subjected to varimax rotation. The first five of
sessive-compulsive personality disorder, but five of the the rotated factors, which together accounted for 51%
seven also had an axis I diagnosis of obsessive-compul- of the variance, were readily interpretable and re-
sive disorder. This suggests that current axis II criteria tained. Similar factors emerged with five-, six-, and
lead clinicians to confound an axis I syndrome with a seven-factor solutions, but the five described here
personality disorder.) proved the most clinically coherent.

Am J Psychiatry 156:2, February 1999 277

REVISING AND ASSESSING AXIS II, PART II

TABLE 4. SWAP-200a Items That Best Describe Patients in TABLE 5. SWAP-200a Items That Best Describe Patients in Ob-
Paranoid Personality Disorder Diagnostic Category (Q-factor 4) sessional Personality Disorder Diagnostic Category (Q-factor 5)
Factor Factor
Itemb Score Itemb Score
Tends to hold grudges; may dwell on insults or slights Tends to be conscientious and responsible. 3.40508
for long periods. 3.60557 Is articulate; can express self well in words. 3.02961
Tends to feel misunderstood, mistreated, or victimized. 3.22667 Has moral and ethical standards and strives to live up
Is quick to assume that others wish to harm or take ad- to them. 3.00354
vantage of him/her; tends to perceive malevolent in- Is able to use his/her talents, abilities, and energy ef-
tentions in others’ words and actions. 3.07566 fectively and productively. 2.45144
Tends to express intense and inappropriate anger, out Enjoys challenges; takes pleasure in accomplishing
of proportion to the situation at hand. 2.77288 things. 2.39760
Tends to be critical of others. 2.59343 Tends to see self as logical and rational, uninfluenced
Tends to get into power struggles. 2.43265 by emotion; prefers to operate as if emotions were ir-
Tends to be angry or hostile (whether consciously or relevant or inconsequential. 2.19796
unconsciously). 2.40625 Is excessively devoted to work and productivity, to the
Tends to see certain others as “all bad” and loses the detriment of leisure and relationships. 2.01831
capacity to perceive any positive qualities the person Tends to be controlling. 1.96135
may have. 2.37861 Is able to find meaning and satisfaction in the pursuit of
Tends to be self-righteous or moralistic. 2.24888 long-term goals and ambitions. 1.95723
Tends to react to criticism with feelings of rage Appreciates and responds to humor. 1.94341
or humiliation. 2.19317 Tends to be inhibited or constricted; has difficulty al-
Tends to blame others for own failures or shortcom- lowing self to acknowledge or express wishes and
ings; tends to believe his/her problems are caused by impulses. 1.89704
external factors. 2.14574 Is able to assert him/herself effectively and appropri-
Tends to be oppositional, contrary, or quick ately when necessary. 1.86883
to disagree. 2.07730 Tends to think in abstract and intellectualized terms,
Tends to see own unacceptable feelings or impulses in even in matters of personal import. 1.83111
other people instead of in him/herself. 2.07543 Has difficulty acknowledging or expressing anger. 1.79087
Tends to become irrational when strong emotions are Tends to be competitive with others (whether con-
stirred up; may show a noticeable decline from cus- sciously or unconsciously). 1.75738
tomary level of functioning. 1.93669 Expects self to be “perfect” (e.g., in appearance,
Tends to “catastrophize”; is prone to see problems as achievements, performance, etc.). 1.75089
disastrous, unsolvable, etc. 1.82261 Tends to elicit liking in others. 1.69161
Tends to elicit dislike or animosity in others. 1.77967 Tends to be overly concerned with rules, procedures,
Emotions tend to spiral out of control, leading to ex- order, organization, schedules, etc. 1.58575
tremes of anxiety, sadness, rage, excitement, etc. 1.72575 Has the capacity to recognize alternative viewpoints,
Has difficulty making sense of other people’s behavior; even in matters that stir up strong feelings. 1.58199
often misunderstands, misinterprets, or is confused Is psychologically insightful; is able to understand self
by others’ actions and reactions. 1.52516 and others in subtle and sophisticated ways. 1.56873
Tends to be controlling. 1.49430 Tends to be self-critical; sets unrealistically high stan-
Tends to elicit extreme reactions or stir up strong feel- dards for self and is intolerant of own human defects. 1.55630
ings in others. 1.44136 Has difficulty allowing self to experience strong, plea-
Tends to avoid confiding in others for fear of betrayal; surable emotions (e.g., excitement, joy, pride). 1.53097
expects things he/she says or does will be used a Shedler-Westen Assessment Procedure—200.
against him/her. 1.40909 b Items listed in descending order of diagnostic import.
Reasoning processes or perceptual experiences seem
odd and idiosyncratic (e.g., may make seemingly ar-
bitrary inferences; may see hidden messages or spe- ited about pursuing goals or successes (their aspira-
cial meanings in ordinary events). 1.38903
Perception of reality can become grossly impaired un- tions or achievements are below their potential).
der stress (e.g., may become delusional). 1.37377 The second subfactor, which we labeled dysphoric:
a Shedler-Westen Assessment Procedure—200. high-functioning neurotic, was characterized by many
b Items listed in descending order of diagnostic import. SWAP-200 statements indicating psychological
strengths, mixed with items indicating chronic dyspho-
ria. The patients’ strengths included being articulate,
Although we will provide only brief descriptions of
having high moral and ethical standards, being em-
these subfactors, they represent important subcatego- pathic, appreciating and responding to humor, being
ries within the dysphoric category. We labeled the first conscientious and responsible, being psychologically
subgroup dysphoric: avoidant. Patients in this category insightful, tending to elicit liking in others, having the
were characterized by SWAP-200 statements indicat- capacity to recognize alternative viewpoints even in
ing (in descending order of importance) that they are matters that stir up strong feelings, being able to hear
shy or reserved, avoid social situations because of fear and benefit from information that is emotionally
of embarrassment, lack social skills, are inhibited or threatening, and being able to sustain a meaningful
constricted, are passive and unassertive, lack close love relationship characterized by genuine intimacy
friendships and relationships, feel like outcasts or out- and caring. Mixed with these positive items were
siders, have difficulty allowing themselves to experi- SWAP-200 items indicating a tendency to blame them-
ence strong pleasurable emotions, feel inadequate or selves or feel responsible for bad things that happen; to
inferior, feel ashamed or embarrassed, and are inhib- feel guilty; to seek out or create relationships in which

278 Am J Psychiatry 156:2, February 1999

 DREW WESTEN AND JONATHAN SHEDLER

TABLE 6. SWAP-200a Items That Best Describe Patients in TABLE 7. SWAP-200a Items That Best Describe Patients in Nar-
Histrionic Personality Disorder Diagnostic Category (Q-factor 6) cissistic Personality Disorder Diagnostic Category (Q-factor 7)
Factor Factor
Itemb Score Itemb Score
Tends to be overly needy or dependent; requires ex- Has fantasies of unlimited success, power, beauty, tal-
cessive reassurance or approval. 3.03132 ent, brilliance, etc. 3.58691
Tends to become attached quickly or intensely; devel- Appears to feel privileged and entitled; expects prefer-
ops feelings, expectations, etc. that are not war- ential treatment. 3.38561
ranted by the history or context of the relationship. 2.97263 Has an exaggerated sense of self-importance. 2.98795
Tends to become attached to, or romantically inter- Seems to treat others primarily as an audience to wit-
ested in, people who are emotionally unavailable. 2.83882 ness own importance, brilliance, beauty, etc. 2.87191
Tends to be suggestible or easily influenced. 2.60990 Seeks to be the center of attention. 2.49405
Tends to be overly sexually seductive or provocative, Expects self to be “perfect” (e.g., in appearance,
whether consciously or unconsciously (e.g., may be achievements, performance, etc.). 2.41758
inappropriately flirtatious, preoccupied with sexual Tends to be arrogant, haughty, or dismissive. 2.40234
conquest, prone to “lead people on,” etc.). 2.58365 Fantasizes about finding ideal, perfect love. 2.38475
Expresses emotion in exaggerated and theatrical Tends to think others are envious of him/her. 2.35242
ways. 2.57185 Tends to feel envious. 2.30673
Fantasizes about finding ideal, perfect love. 2.49440 Tends to be competitive with others (whether con-
Is unable to soothe or comfort self when distressed; re- sciously or unconsciously). 2.29974
quires involvement of another person to help regu- Tends to believe he/she can only be appreciated by, or
late affect. 2.13633 should only associate with, people who are high-sta-
Emotions tend to spiral out of control, leading to ex- tus, superior, or otherwise “special.” 2.20950
tremes of anxiety, sadness, rage, excitement, etc. 2.13556 Lacks close friendships and relationships. 1.93136
Tends to fear he/she will be rejected or abandoned by Tends to feel he/she is not his/her true self with others;
those who are emotionally significant. 2.11823 tends to feel false or fraudulent. 1.73900
Tends to use his/her physical attractiveness to an ex- Tends to use his/her physical attractiveness to an ex-
cessive degree to gain attention or notice. 1.94586 cessive degree to gain attention or notice. 1.67349
Tends to be anxious. 1.92664 Tends to feel life has no meaning. 1.55898
Tends to choose sexual or romantic partners who Has a disturbed or distorted body image; sees self as
seem inappropriate in terms of age, status (e.g., so- unattractive, grotesque, disgusting, etc. 1.43804
cial, economic, intellectual), etc. 1.79183 Tends to feel empty or bored. 1.41987
Tends to develop somatic symptoms in response to Tends to react to criticism with feelings of rage
stress or conflict (e.g., headache, backache, abdom- or humiliation. 1.40427
inal pain, asthma, etc.). 1.78542 Appears afraid of commitment to a long-term
Perceptions seem glib, global, and impressionistic; love relationship. 1.39740
has difficulty focusing on specific details. 1.75259 Tends to feel like an outcast or outsider; feels as
Seeks to be the center of attention. 1.67221 if he/she does not truly belong. 1.37469
Emotions tend to change rapidly and unpredictably. 1.66518 Has little empathy; seems unable to understand or re-
Tends to become irrational when strong emotions are spond to others’ needs and feelings unless they co-
stirred up; may show a noticeable decline from cus- incide with his/her own. 1.30599
tomary level of functioning. 1.60671 Tends to seek power or influence over others (whether
Appears to fear being alone; may go to great lengths to in beneficial or destructive ways). 1.26300
avoid being alone. 1.49096 Tends to be self-critical; sets unrealistically high stan-
Tends to get drawn into or remain in relationships in dards for self and is intolerant of own human defects. 1.23000
which he/she is emotionally or physically abused. 1.47383 Feels some important other has a special, almost
a Shedler-Westen Assessment Procedure—200. magical ability to understand his/her innermost
b Items listed in descending order of diagnostic import. thoughts and feelings (e.g., may imagine rapport is
so perfect that ordinary efforts at communication are
superfluous). 1.18843
they are in the role of caring for, rescuing, or protecting a Shedler-Westen Assessment Procedure—200.
b Items listed in descending order of diagnostic import.
the other; to feel unhappy, depressed, or despondent;
to fear they will be rejected or abandoned; to be self-
critical; to be anxious; and to be insufficiently con- needy and dependent, and a tendency to engage in self-
cerned with meeting their own needs. mutilating behavior.
The third subfactor, which included many patients The fourth subfactor, labeled dysphoric: dependent-
currently diagnosed as borderline, was labeled dys- masochistic, includes patients who appear to be much
phoric: emotionally dysregulated. These patients were more disturbed than those in the current axis II depen-
characterized by SWAP-200 statements describing dent category. These patients tend to get drawn into or
remain in relationships in which they are emotionally
emotions that spiral out of control, struggles with gen-
or physically abused; are ingratiating or submissive;
uine suicidal wishes, an inability to soothe or comfort become attached quickly or intensely (develop feelings
themselves when distressed, a tendency to feel life has or expectations that are not warranted by the history
no meaning, a tendency to make repeated suicidal or context of the relationship); are suggestible or easily
threats or gestures, a tendency to “catastrophize” (see influenced; become attached to, or romantically inter-
problems as disastrous and unsolvable), a tendency to ested in, people who are emotionally unavailable; are
become irrational when strong emotions are stirred up, overly needy or dependent; fear being alone; fear they
a tendency to feel empty or bored, a tendency to be will be rejected or abandoned; express aggression in

Am J Psychiatry 156:2, February 1999 279

REVISING AND ASSESSING AXIS II, PART II

TABLE 8. Correlations (r) Between Q-Factor Scores From the SWAP-200a and Clinicians’ Ratings (N=496)
Q-Factorb
Clinician Rating Dysphoric Antisocial Schizoid Paranoid Obsessional Histrionic Narcissistic
Paranoid –0.27 0.30 0.18 0.53 –0.27 –0.21 0.05
Schizoid –0.04 –0.08 0.68 0.04 –0.20 –0.41 –0.03
Schizotypal –0.15 0.01 0.52 0.18 –0.36 –0.24 –0.07
Antisocial –0.48 0.70 –0.13 –0.01 –0.35 0.01 0.06
Borderline 0.02 0.26 –0.23 0.17 –0.48 0.26 –0.07
Histrionic –0.05 0.11 –0.40 0.03 –0.25 0.58 0.09
Narcissistic –0.38 0.44 –0.29 0.18 –0.15 0.21 0.51
Avoidant 0.40 –0.28 0.44 –0.11 –0.06 –0.30 –0.02
Dependent 0.38 –0.16 –0.01 –0.10 –0.17 0.31 –0.16
Obsessive-compulsive 0.07 –0.30 0.02 0.10 0.32 –0.22 0.01
Depressive 0.49 –0.19 0.04 –0.01 –0.13 –0.01 –0.05
Passive-aggressive –0.04 0.29 0.04 0.08 –0.20 0.04 –0.07
Self-defeating 0.30 –0.01 0.00 –0.02 –0.25 0.06 –0.15
Sadistic –0.48 0.56 –0.19 0.19 –0.23 –0.08 0.15
Global Assessment of
Functioning Scalec –0.13 –0.13 –0.33 –0.16 0.52 0.03 0.23
a Shedler-Westen Assessment Procedure—200.
b For all correlation coefficients of 0.15 or greater, p<0.001.
c Numbers in bold and italic indicate convergent validity coefficients.

passive and indirect ways; and lack a stable image of Q-scores on our antisocial Q-factor also received high
who they are or would like to become. ratings on an independent measure of antisocial per-
The final subfactor was labeled dysphoric: hostile- sonality. Table 8 presents correlations for the primary
externalizing and contained patients who were hostile seven Q-factors, and table 9 presents correlations for
and prone to blame others for their difficulties, with the dysphoric subfactors.
passive-aggressive features. The SWAP-200 statements As tables 8 and 9 show, Q-scores for categories that
described a tendency to get into power struggles; to be resemble current axis II diagnostic (i.e., antisocial,
angry or hostile; to blame others for their own failures schizoid, paranoid, histrionic, and narcissistic) have
or shortcomings; to feel misunderstood, mistreated, or uniformly high correlations with clinicians’ ratings for
victimized; to be critical of others; to be conflicted about those axis II categories, indicating convergent validity.
authority (to feel they must submit, rebel against, win Of equal importance, the Q-scores had low correla-
over, defeat, and so on); to hold grudges; to express ag- tions with clinicians’ ratings for unrelated categories,
gression in passive and indirect ways; to be opposi- indicating strong discriminant validity. Note, for ex-
tional and contrary; and to feel helpless or powerless. ample, the distinctiveness of the narcissistic Q-score,
which correlates highly with clinician ratings of narcis-
Validity of the Empirically Derived Taxonomy sistic personality disorder (r=0.51) and with nothing
else. We are not aware of other personality disorder
We created a composite Q-sort description of the pa- measures that have been able to distinguish narcissistic
tients in each Q-factor, to serve as a diagnostic proto- personality disorder so clearly from other cluster B dis-
type or template for the Q-factor (see part I). We then orders, particularly antisocial.
computed the correlation between each patient’s In addition, tables 8 and 9 help us better understand
SWAP-200 description and each diagnostic prototype, the new diagnostic categories (e.g., dysphoric, and the
to gauge the “match” between each patient and each dysphoric subfactors), relative to the familiar axis II di-
Q-factor. We will refer to these correlation coefficients agnostic categories. As can be seen, patients in the dys-
as Q-scores. Thus, each patient received 12 Q-scores, phoric Q-factor have avoidant, dependent, and depres-
one for each of the seven primary Q-factors and one sive features (and do not have antisocial or narcissistic
for each of the five dysphoric subfactors (e.g., a given features); patients in the dysphoric: emotionally dys-
patient might have a Q-score of 0.44 for the dysphoric regulated subfactor have borderline features.
Q-factor, –0.06 for the antisocial-psychopathic Q-fac- Finally, tables 8 and 9 list the correlations between
tor, 0.11 for the schizoid Q-factor, and so on). Q-scores and Global Assessment of Functioning Scale
As an initial test of the validity of the new empiri- ratings, which were provided by the reporting clini-
cally derived personality disorder typology, we exam- cians. The pattern of correlations indicates that the Q-
ined the relations between Q-scores and clinicians’ factors form a hierarchy of pathology, with roughly
ratings of the extent to which patients met current three levels. In order of increasing pathology, these are
personality disorder criteria (7-point rating scale; 1= 1) narcissistic, obsessional, and dysphoric: high-func-
“not at all,” 4=“has some features,” and 7=“fully tioning neurotic; 2) paranoid, antisocial, histrionic,
meets criteria”). Because many of the Q-factors are dysphoric, dysphoric: avoidant, dysphoric: dependent-
similar to current axis II diagnostic categories, this al- masochistic, and dysphoric: hostile-externalizing; and
lowed us to verify, for example, that patients with high 3) schizoid and dysphoric: emotionally dysregulated.

280 Am J Psychiatry 156:2, February 1999

 DREW WESTEN AND JONATHAN SHEDLER

TABLE 9. Correlations (r) Between Dysphoric Subfactor Scores and Clinicians’ Ratings (N=496)
Dysphoric Subfactora
High-Functioning Emotionally Dependent- Hostile-
Clinician Rating Avoidant Neurotic Dysregulated Masochistic Externalizing
Paranoid –0.16 –0.40 0.22 –0.28 0.37
Schizoid 0.41 –0.42 0.08 –0.32 0.07
Schizotypal 0.18 –0.45 0.21 –0.24 0.03
Antisocial –0.47 –0.44 –0.04 0.00 0.25
Borderline –0.31 –0.15 0.51 0.31 –0.03
Histrionic –0.39 0.03 0.20 0.42 –0.15
Narcissistic –0.49 –0.23 –0.02 0.02 0.18
Avoidant 0.56 –0.03 0.17 0.00 –0.04
Dependent 0.15 0.09 0.27 0.47 –0.11
Obsessive-compulsive 0.16 0.14 –0.07 –0.28 –0.08
Depressive 0.26 0.15 0.39 0.23 –0.02
Passive-aggressive –0.10 –0.24 0.07 0.16 0.36
Self-defeating 0.10 –0.01 0.31 0.33 0.03
Sadistic –0.47 –0.38 –0.05 –0.15 0.28
Global Assessment of Functioning
Scaleb –0.12 0.37 –0.47 –0.13 –0.15
a For all correlation coefficients of 0.15 or greater, p<0.001.
b Numbers in bold and italic indicate convergent validity coefficients.

DISCUSSION gories. This suggests that our efforts to develop catego-

ries with minimal overlap may have been successful;
The SWAP-200 allows clinicians to provide detailed, definitive findings in that regard await replication.
clinically rich descriptions of patients in a systematic A major finding of the study was the emergence of a
and quantifiable form. These quantified clinical obser- dysphoric Q-factor, which included roughly 20% of
vations may be useful for refining axis II because they the patient group and was by far the largest category,
generate a classification system that is both empirically despite its lack of recognition in DSM-IV. The Q-factor
grounded and faithful to clinical experience. We iden- included many patients now diagnosed as dependent,
tified seven Q-factors or diagnostic categories (or 11, if avoidant, depressive, self-defeating, and borderline.
the dysphoric category is divided into five subcatego- The fact that these patients formed a distinct diagnos-
ries), some of which resemble current axis II categories tic category suggests that axis II may have overfo-
and some of which do not. The psychological features cused on the ways such patients are socially dysfunc-
associated with each category appear to be clinically tional and underfocused on the ways in which they
and theoretically coherent, suggesting that the catego- are in pain. The data support the inclusion of a de-
ries represent meaningful clinical syndromes. In addi- pressive/dysphoric personality disorder diagnosis in
tion, the classification system avoids many of the con- DSM-V (see reference 19). We also identified distinct
ceptual and empirical problems associated with the subcategories within the dysphoric category. The sub-
current axis II taxonomy (discussed in part I), includ- categories encourage a functional approach to under-
ing 1) unacceptably high comorbidity of personality standing personality disorders because they represent
disorder diagnoses, 2) artificially dichotomizing con- not only different triggering conditions for distress
tinuous variables (diagnostic criteria) into present/ab- (e.g., social interaction versus abandonment) but also
sent, 3) assuming that personality pathology is cate- different styles of regulating painful affect. Thus,
gorical, 4) failing to weight criteria that differ in the some dysphoric patients respond to pain by self-muti-
degree to which they are diagnostic, 5) neglecting lation or desperately seeking others for soothing (dys-
healthy aspects of functioning, and 6) lack of fidelity to phoric: emotionally dysregulated), others become
findings from cluster or factor analytic studies. needy and dependent (dysphoric: dependent-maso-
Q-scores (which measure the extent to which pa- chistic), others avoid interactions that may cause anx-
tients have the characteristics of each Q-factor) corre- iety or feelings of rejection and inadequacy (dysphoric:
lated in meaningful ways with clinician ratings of the avoidant), and so on.
extent to which patients met current axis II diagnostic A second, and clinically sensible, finding was the
criteria. Thus, Q-scores correlated highly with clini- identification of a revised schizoid category that in-
cian ratings for similar axis II diagnoses and did not cludes many patients currently diagnosed as schizoid,
correlate with ratings for unrelated diagnoses. The pat- avoidant, or schizotypal—three categories that are no-
tern of correlations suggests strong convergent and dis- toriously difficult to distinguish. The data suggest that
criminant validity. Indeed, the findings were much these categories are difficult to distinguish because
stronger than those observed for the current axis II cat- they are not empirically distinct. (The absence of a
egories (see part I), which is striking since the criterion schizotypal category is worth comment, especially in
measures (axis II ratings) were based on current cate- light of research evidence of the genetic basis of schizo-

Am J Psychiatry 156:2, February 1999 281

REVISING AND ASSESSING AXIS II, PART II

typy. We believe a schizotypal category did not emerge Implications

because schizotypy is not a personality disorder [de-
fined by a unique constellation of personality pro- Data such as these need to be replicated and, in partic-
cesses] but a clinical syndrome like schizophrenia de- ular, need to be replicated in a broader patient group that
fined by a single trait [low-grade thought disorder] does not include only patients preselected by clinicians to
that might be better diagnosed on axis I. We address meet current axis II personality disorder criteria. Instead,
this issue in a paper in preparation, in which we isolate the same procedures could be undertaken in a large sam-
a subclinical thought disorder factor through factor ple of randomly selected patients being treated for mal-
analysis, that predicts genetic history of psychosis in adaptive personality patterns, who may or may not meet
first-degree relatives and appears to be taxonic.) current axis II criteria for a personality disorder. The pa-
tients’ SWAP-200 descriptions could then be subjected to
The Q-analysis also identified a revised histrionic di-
Q-factor analysis and subsequently to taxometric analy-
agnostic category that included many items currently
sis (23), to determine what diagnostic categories emerge
in the DSM description of histrionic personality disor-
in a broader sample and whether those categories are
der along with several items associated with borderline
true taxons or are better understood as dimensions (con-
personality disorder—a category that shows high co-
tinua). An advantage of the Q-factors is that they can be
morbidity with histrionic personality disorder in all
treated as dimensions, categories, or both.
studies of which we are aware. These findings suggest
The data also suggest that axis II diagnostic criteria
that some patients currently diagnosed with borderline
need not be confined to manifestly observable symp-
personality disorder may be better classified within the
toms but should also include criteria that describe per-
dysphoric spectrum (especially in the dysphoric: emo-
sonality dynamics underlying these manifest symp-
tionally dysregulated category, and to a lesser extent in
toms. For example, if defensive processes can be
the dysphoric: dependent-masochistic subcategory),
assessed reliably using jargon-free items that do not re-
while others may be better diagnosed as histrionic. An
quire a commitment to a particular theory, then those
important distinction between them is that dysphoric items should be integrated into diagnostic profiles
patients’ affective intensity is highly ego-dystonic, where they prove diagnostically useful and not rele-
whereas histrionic patients’ affective intensity is syn- gated to an appendix (16 and 24). The use of such cri-
tonic. We have now replicated this finding in two inde- teria would also increase the clinical usefulness of
pendent patient samples (17). DSM, since many SWAP-200 items describe psycho-
The data also support arguments for a dimensional logical processes of this sort that will be addressed in
system for diagnosing personality disorders, either in treatment or psychological strengths that clinicians
place of the present categorical system or, probably will draw upon in treatment. In contrast, current axis
more useful, in combination with it (20, 21). A dimen- II criteria provide little insight into many treatment-
sional approach would treat personality pathology as a relevant issues (e.g., the function of psychological
continuum, not as a present/absent dichotomy. Rele- symptoms, the processes that maintain them, the ac-
vant to this is the finding that not all Q-factors were companying affect, the manner of regulating distress).
comparable with respect to level of functioning. For Perhaps the most important feature of the SWAP-
example, patients in the obsessional Q-factor and the 200 method is its potential to bridge the gap that too
dysphoric: high-functioning neurotic subfactor were often separates clinical and empirical approaches to
considerably healthier (e.g., had higher Global Assess- personality pathology. The categories and criteria that
ment of Functioning Scale scores) than patients in emerge through use of the SWAP-200 procedure have
other categories. Their pathology might better be con- clinical validity because they are derived from clinical
ceptualized in terms of neurotic styles (17) than in observation. In addition, they meet or exceed requisite
terms of personality disorders. We suspect that in a less standards for psychometric validity. With the SWAP-
constrained patient group, where clinicians described 200, clinicians do not need to administer a special
patients with personality pathology but not necessarily questionnaire or structured interview to obtain mean-
axis II disorders, other neurotic styles would appear, ingful clinical or research data. They just need to draw
such as the hysterical style (17) that influenced the appropriate and reasonably straightforward inferences
DSM description of histrionic personality disorder but from the clinical data already available to them—
is much more benign. Whether these neurotic styles are which is exactly what well-trained clinicians should be
simply less severe versions of personality disorders, or able to do—and express those inferences by using the
represent categories sui generis, is a topic worth em- standard vocabulary of the SWAP-200. Our studies
pirical attention—especially since over 80% of expe- show that clinicians are quite capable of drawing reli-
rienced psychologists and psychiatrists report treating able inferences from clinical data, as evidenced by
patients for personality pathology not severe enough strong correlations between SWAP-200 descriptions of
to warrant an axis II diagnosis [21] and over 60% of the same patient by independent clinicians (17) and by
patients being treated for maladaptive personality meaningful patterns of convergent and discriminant
patterns cannot be diagnosed on axis II (22). A Q-an- validity coefficients.
alytic study of a group of these patients is currently The SWAP-200 helps bridge the gap between clinical
underway. and research approaches in another way, by providing

282 Am J Psychiatry 156:2, February 1999

 DREW WESTEN AND JONATHAN SHEDLER

not only diagnostic categories and dimensions, but FIGURE 1. A Diagnostic Profile for Mr. N
also narrative case descriptions of patients. To present
case formulations, clinicians need only list the 18 to 30
SWAP-200 items placed in the highest categories and
use these statements to “anchor” their clinical infer-
ences and formulations (see reference 25). When this
procedure is used, case formulation and diagnosis flow
from the same procedure and are not the unrelated en-
terprises that they now tend to be (7).

A Case Example: Using the SWAP-200 to Diagnose

an Individual Patient

We now illustrate the use of the SWAP-200 to diag-

nose an individual patient and also provide a narrative
case study. The patient, who we will call Mr. N, was
chosen from among the 496 patients in this study. Mr.
N is a 48-year-old white man with a college education,
seen for nine psychotherapy sessions at the time the
treating clinician described him using the SWAP-200.
The treating clinician gave him an axis I diagnosis of
adjustment disorder and an axis II diagnosis of narcis-
sistic personality disorder. He is relatively high func- tant other has a special, almost magical ability to un-
tioning, with a Global Assessment of Functioning Scale derstand his innermost thoughts and feelings (e.g., he
score of 65. The clinician reported no noteworthy may imagine rapport is so perfect that ordinary efforts
childhood traumas, although he rated Mr. N’s relation- at communication are superfluous).
ship with his father as very poor. Genetic history is un- He tends to be angry or hostile (whether consciously
remarkable. or unconsciously), tends to be controlling, and tends to
Q-score profile. Figure 1 presents Mr. N’s Q-score be conflicted about authority (e.g., may feel he must
profile, showing the match between Mr. N’s SWAP-200 submit, rebel against, win over, defeat). He tends to ex-
description and each of the seven primary Q-factors. press aggression in passive and indirect ways (e.g., may
For ease of interpretation, we have transformed the raw make mistakes, procrastinate, forget, become sulky)
Q-scores (which are correlation coefficients) into T and to think in abstract and intellectualized terms,
scores, which have a mean of 50 and standard deviation even in matters of personal import. He repeatedly con-
of 10. (T scores are the metric used by the MMPI and vinces others of his commitment to change, only to re-
many other psychological tests and are familiar to most vert to his previous maladaptive behavior (i.e., he con-
personality researchers and many clinicians.) Our data vinces people that “this time is really different”).
suggest that a T score of 70 or higher (two standard de- Mr. N uses his physical attractiveness to an excessive
viations above the mean) is the appropriate cutoff for degree to gain attention and notice; tends to be overly
making a categorical personality disorder diagnosis us- sexually seductive or provocative, whether consciously
ing the diagnostic categories derived by Q-analysis or unconsciously (e.g., may be inappropriately flirta-
(which are more distinct than those that describe cur- tious, preoccupied with sexual conquest, prone to
rent diagnoses, as in part I). Thus, Mr. N’s Q-score pro- “lead people on”); tends to be hostile toward members
file indicates a narcissistic personality disorder with ob- of the opposite sex, whether consciously or uncon-
sessional features. He also appears to have some sciously; appears afraid of commitment to a long-term
histrionic qualities. love relationship; has an active and satisfying sex life;
Narrative description. The reporting clinician placed and fantasizes about finding ideal, perfect love.
the following SWAP-200 items in the top three (most Along with his pathology, Mr. N has considerable
descriptive) categories. The items are reprinted nearly psychological strengths: He is energetic and outgoing;
verbatim, with only minor grammatical changes to aid tends to elicit liking in others; is articulate; appreciates
the flow of the text. and responds to humor; appears comfortable and at
Mr. N has an exaggerated sense of self-importance; ease in social situations; is creative and able to ap-
feels privileged and entitled; believes he can only be ap- proach problems in novel ways; can assert himself ef-
preciated by, or should only associate with, people fectively and appropriately when necessary; and ap-
who are high-status, superior, or otherwise “special”; pears to have come to terms with painful experiences
fantasizes about unlimited success, power, beauty, tal- from the past, having found meaning in, and grown
ent, brilliance, and so on; treats others primarily as an from, such experiences.
audience to witness his importance, brilliance, beauty, Case formulation. From this configuration of per-
etc; seeks to be the center of attention; tends to be ar- sonality characteristics, we draw the following clinical
rogant, haughty, and dismissive; and feels an impor- inferences, or hypotheses. Mr. N is a high-functioning

Am J Psychiatry 156:2, February 1999 283

REVISING AND ASSESSING AXIS II, PART II

narcissistic character. He can be charming and likable for classifying and assessing personality for clinical
and uses his charm to win admiration and affection. At purposes should, we believe, encompass this human
the same time, he is self-centered and entitled and val- complexity, not describe caricatures.
ues others primarily to the extent that they bolster his Third, the SWAP-200 can be used not only to clas-
grandiose (but fragile) view of himself (e.g., by offering sify patients into categories, such as narcissistic per-
admiration or witnessing his magnificence). His rela- sonality disorder, but to make fine-grained distinctions
tionships probably begin with promise, only to sour among patients within a diagnostic category. For ex-
with time. These personality dynamics find particular ample, we note that Mr. N has a capacity for genuine
expression in Mr. N’s relations with women. We sus- insight and growth and suspect that his prognosis is far
pect he is a womanizer who leaves victims in his wake, more favorable than that of many other patients with
because he is charming and leads women on but is un- narcissistic personality disorder, particularly those
able to sustain a truly meaningful relationship charac- with more antisocial features. With the SWAP-200,
terized by mutual empathy, caring, and sharing. He this is a testable hypothesis.
cannot do so because at core he seeks someone whose
true role is to help regulate his self-esteem, e.g., by un- Toward DSM-V
derstanding him perfectly, admiring his perfection, and
being perfect herself. He is angry and subtly devaluing Thus far, we have avoided the thorny question of how
toward women, who do not quite fulfill these wishes. the findings of this study, if they prove replicable, could
At age 48, his fantasies about ideal, perfect love are in- be used to revise axis II. Using the actual SWAP-200
creasingly difficult to sustain. He is probably confused procedure would be essential for diagnosis in research
and pained by his repeated failed relationships, and contexts and would also make sense in certain clinical
this may be what has brought him to treatment. situations, such as forensic evaluations, or in cases
If our inferences are correct, Mr. N is likely to express where the diagnostic picture is unclear. In these cases, a
these issues in the therapeutic relationship in a variety of Q-score profile (figure 1) could provide both dimen-
ways. He might seek out a therapist who he can see as sional and categorical diagnosis (where categorical diag-
special and superior like himself, who will share in his noses are made when Q-scores exceed a critical cutoff
perfection and understand him perfectly; and/or he may point, such as two standard deviations above the mean).
demean and devalue the therapist, who must ultimately In daily clinical practice, routine use of the SWAP-
disappoint. Mr. N’s tendency to intellectualize will pres- 200 would be impractical and fortunately is unneces-
ent difficulties, since he may treat therapeutic insights as sary. The simplest way to revise axis II, which would
“theories” to ponder, without the personal relevance preserve much of its familiar format and hence be
and affective charge that leads to change. The fact that readily used by clinicians, would be to replace the cur-
he has found meaning in past painful experiences and rent approach with a prototype matching procedure
grown from them, however, along with his many that yields both dimensional and categorical diag-
strengths, increases our confidence that he will use ther- noses. The most diagnostic or important items for
apy effectively. each Q-factor could serve as criteria sets. To maxi-
mize internal consistency and minimize comorbidity,
Comments on the Case Formulation the number of criteria per disorder would need to be
greater than the current seven to 10 criteria per disor-
We leave to readers to judge the merits of the SWAP- der. Thus, the prototype for each disorder might in-
200 as an assessment tool, relative to current instru- clude the top 18 SWAP-200 items (this is the number
ments intended to assess personality disorders. Three of SWAP-200 items in the two “most descriptive” cat-
points, however, are worth noting. First, with the egories, 6 and 7), or each category could include as
SWAP-200 approach, diagnosis and case formulation many items as necessary to achieve a coefficient alpha
are part of the same process. The SWAP-200 provides >0.80 for the item set (i.e., high internal consistency).
both a Q-scores profile for diagnostic purposes (figure Thus, the criteria sets would resemble the listing of
1) and a narrative description useful for case formula- items in tables 1 through 7, arranged in descending
tion. Moreover, the standard vocabulary of the SWAP- order of importance.
200 ensures that different clinicians will describe the To make a diagnosis, clinicians could simply rate the
same patient in much the same way once they learn to extent to which a patient matches each of the seven cri-
use the SWAP-200 reliably. Had another clinician de- teria sets (or 11 criteria sets, if the dysphoric category
scribed Mr. N using the SWAP-200, the narrative de- is divided into five subtypes), using, for example, a 0–
scription would have been much the same (since every 7 rating scale (0=the patient has no resemblance to the
word was taken directly from the SWAP-200 item set). diagnostic prototype, 3=the patient has features of the
Second, the constellation of seemingly contradictory disorder, 5=the patient matches the prototype well
qualities embodied in Mr. N’s narrative description— enough to receive a diagnosis, 7=the patient is a rela-
such as the paradoxical combination of several un- tively pure and prototypic example of the disorder). If
pleasant narcissistic qualities and a tendency to elicit such a rating system were used, Mr. N might receive
liking in others—is normative with the SWAP-200 and, the following dimensional diagnoses: dysphoric per-
we believe, normative in human personality. A system sonality disorder, 0; antisocial personality disorder, 1;

284 Am J Psychiatry 156:2, February 1999

 DREW WESTEN AND JONATHAN SHEDLER

schizoid personality disorder, 0; paranoid personality Factor Model of Personality. Edited by Costa P, Widiger T.
disorder, 0; obsessional personality disorder, 3; histri- Washington, DC, American Psychological Association, 1994,
pp 19–39
onic personality disorder, 2; and narcissistic personal-
2. Livesley WJ (ed): The DSM-IV Personality Disorders. New
ity disorder, 5. The categorical diagnosis would be nar- York, Guilford Press, 1995
cissistic personality disorder with obsessional features. 3. Morey LC: Personality disorders in DSM-III and DSM-III-R:
This rating procedure would likely take no more convergence, coverage, and internal consistency. Am J Psy-
than a moment or two of a clinician’s time once the cli- chiatry 1988; 145:573–577
nician became familiar with the diagnostic manual. It 4. Bell E, Jackson D: The structure of personality disorders in
DSM-III. Acta Psychiatr Scand 1992; 85:279–287
would yield diagnoses similar to those currently used
5. Mischel W, Shoda Y: A cognitive-affective system theory of
by clinicians, such as “paranoid personality disorder personality: reconceptualizing situations, dispositions, dy-
with schizoid features.” Whether this approach would namics, and invariance in personality structure. Psychol Rev
render clinical diagnoses more or less reliable is an em- 1995; 102:246–268
pirical question, although we doubt it would prove less 6. Westen D: A clinical-empirical model of personality: life after
reliable than current categorical clinical diagnoses. the Mischelian ice age and the NEO-lithic era. J Personality
1995; 63:495–524
Limitations 7. Westen D: Case formulation and personality diagnosis: two
processes or one? in Making Diagnosis Meaningful. Edited by
Barron J. Washington, DC, American Psychological Associa-
Beyond the potential objections discussed in part I, tion Press, 1998, pp 111–138
this study has a number of limitations, of which we be- 8. Bernstein DP, Useda D, Siever L: Paranoid personality disor-
lieve two are most significant and point the way toward der, in The DSM-IV Personality Disorders. Edited by Livesley
future research. The first is that the research design was WJ. New York, Guilford Press, 1995, pp 45–57
biased to increase the chances of duplicating the current 9. Hare R, Hart S: Commentary on antisocial personality disor-
axis II taxonomy, since our clinicians described only pa- der: the DSM-IV field trial. Ibid, pp 127–134
tients who met current axis II criteria for a personality 10. Block J: Lives Through Time. Berkeley, Calif, Bancroft, 1971
disorder. We deliberately chose this research strategy so 11. Block J: The Q-Sort Method in Personality Assessment and
Psychiatric Research. Palo Alto, Calif, Consulting Psycholo-
we could 1) evaluate the validity of the new approach gists Press, 1978
relative to familiar axis II categories, and 2) be certain 12. Block J, Gjerde P, Block J: Personality antecedents of depres-
that we included patients who spanned the entire spec- sive tendencies in 18-year-olds: a prospective study. J Pers
trum of current, recently recognized, and “under con- Soc Psychol 1991; 60:726–738
sideration” personality disorder syndromes. Given this 13. Colvin R, Block J, Funder D: Overly positive self evaluations
research strategy, it is striking that the Q-factor analysis and personality: negative implications for mental health. J
Pers Soc Psychol 1995; 68:1152–1162
did not simply replicate the current taxonomy, but in-
14. John O, Robins RW: Accuracy and bias in self-perception: in-
stead suggested an alternative taxonomy. dividual differences in self-enhancement and the role of nar-
Second, this study relied exclusively on data from cissism. J Pers Social Psychol 1994; 66:206–219
only one source, clinicians. Future research should in- 15. Shedler J, Block J: Adolescent drug use and psychological
clude other data sources (e.g., self-reports from pa- health: a longitudinal inquiry. Am Psychol 1990; 45:612–630
tients, data from informants) to triangulate on the con- 16. Westen D, Muderrisoglu S, Fowler C, Shedler J, Koren D: Af-
structs of interest, and to evaluate whether other data fect regulation and affective experience: individual differ-
ences, group differences, and measurement using a Q-sort
sources support the empirically derived taxonomy. We procedure. J Consult Clin Psychol 1997; 65:429–439
would not expect total concordance between data 17. Shedler J, Westen D: Refining the measurement of axis II: a
sources, since we do not believe that patients can self- Q-sort procedure for assessing personality pathology. As-
report about many important personality processes. sessment 1998; 5:335–355
Nevertheless, some concordance seems likely, and 18. Shapiro D: Neurotic Styles. New York, Basic Books, 1965
comparable findings across assessment methods would 19. Phillips K, Hirschfeld RMA, Shea MT, Gunderson JG: Depres-
increase our confidence in the obtained findings. sive personality disorder, in The DSM-IV Personality Disor-
ders. Edited by Livesley WJ. New York, Guilford Press, 1995,
The problems inherent in relying on a single data pp 287–302
source (in this case, clinicians) are by no means unique 20. Livesley WJ: Preface. Ibid, pp v–x
to this study, since most published studies of personality 21. Westen D: Divergences between clinical and research meth-
and personality disorders rely exclusively on a single ods for assessing personality disorders: implications for re-
data source, usually an interviewer’s judgment after a search and the evolution of axis II. Am J Psychiatry 1997; 154:
brief clinical interview, or self-report data, both of 895–903
which ultimately rely on patients’ answers to direct 22. Westen D, Arkowitz-Westen L: Limitations of axis II in diag-
nosing personality pathology in clinical practice. Am J Psychi-
questions. We believe the quantified clinical data re- atry 1998; 155:1767–1771
ported here, obtained from clinicians who have worked 23. Meehl P: Bootstraps taxometrics: Solving the classification
with patients longitudinally, are at least as sound as problem in psychopathology. Am Psychol 1995; 50:266–275
data obtained through the more usual methods. 24. Vaillant G (ed): Ego Mechanisms of Defense: A Guide for Cli-
nicians and Researchers. Washington, DC, American Psychi-
REFERENCES atric Press, 1992
25. Jones E, Windholz M: The psychoanalytic case study: toward
1. Widiger T, Frances A: Towards a dimensional model for the a method for systematic inquiry. J Am Psychoanal Assoc
personality disorders, in Personality Disorders and the Five- 1990; 38:985–1012

Am J Psychiatry 156:2, February 1999 285