CODEPENDENCE, NARCISSISM, AND CHILDHOlOD TRAUMA

HARVEY J . IRWIN
University of New England, Australia

Codependence has been held to be a product of living in a household with

an alcoholic parent or, more generally, an outcome of childhood abuse.
Codependent traits also have been proposed to have a complementary
developmental relationship with narcissism. Australian adults ( N = 190)
were administered the Codependency Inventory, the Spann-Fischer Code-
pendency Scale, the Narcissistic Personality Inventory, the Narcissistic Per-
sonality Disorder scale, the Children of Alcoholics Screening Test, and the
Survey of Traumatic Childhood Events. Results indicated that codependence
is not predictable by childhood trauma, and although a relationship between
codependence and narcissism was established, it was rather more complex
than that anticipated by the literature. These findings substantially weaken
the theoretical underpinnings of the concept of codependmency, and due
caution should be exercised in psychotherapeuticapplications of the concept.

The concept of codependence currently is utilized very widely in the area of mental
health, particularly in the context of self-help literature and support groups. The objec-
tive of the present study was to subject to empirical scrutiny iwo proposals made in
the codependence literature.
The term “codependence” has popularist origins and in part as a consequence of
this, definitions of the concept tend t o be diverse and lacking in rigor (Gomberg, 1989;
Krestan & Bepko, 1990; Myer, Peterson, & Stoffel-Rosales, 1991). In its most recent
usage, codependence concerns a dysfunctional style of relating to others. More specifi-
cally, the codependent person is said to base his or her identiiy on a caretaking role
and assumes an inordinate level of responsibility for the lives, feelings, and problems
of other people (O’Brien & Gaborit, 1992). By investing so heavily in the life of another
person, the codependent tends t o be characterized by low self-worth, an inability to get
his or her own needs met, and a self-identity founded predominantly on external valida-
tion (Schaef, 1986).
Despite its similarity to Horney’s (1 942) notion of morbid dependency, the concept
of codependence has its origins in grass-roots movements associated with alcohol
dependence (Haaken, 1993). Prior t o 1950 the treatment of alcoholism had focused ex-
clusively on the addict, but in 1951 the foundation of Al-Anon by the wives of alcoholics
brought attention to the role of the spouse in reinforcing an alcoholic’s behavior. The
alcoholic’s spouse thus was seen to be taking responsibility foir the alcoholic’s life by
attempting both t o control and to protect the problem drinker The spouse seemingly
derived a sense of identity from this caretaking role. Soon after the term “codependence”
was coined in the late 1970s, the concept became increasingly generalized. First it was
realized that the children of an alcoholic could be similarly affected. Indeed, the 1980s
saw a growing conviction that the spouse of an alcoholic did not really become codepen-
dent in the relationship with the alcoholic but, rather, had developed codependence in
childhood by growing up in the household of an alcoholic parent; later in life the co-
dependent person entered a conjugal relationship with another alcoholic because the
latter fulfilled the codependent’s self-perceived function to take responsibility for the lives

Correspondence should be addressed to Assoc. Prof. H. J . Irwin, Department of Psychology, University

of New England, Armidale, NSW 2351, Australia.

658
 Codependence, Narcissism, and Childhood Trauma 659

of others. More recently, the concept of codependence has been applied beyond the con-
text of alcohol dependence. Codependence now generally is viewed as a product of a
childhood spent in a family with any of a wide range of dysfunctions, which renders
the person prone in later life to forming dysfunctional caretaking relationships with ad-
dictive or compulsive individuals (Prest & Protinsky, 1993).
In the current context, therefore, there is a common view that codependence is not
specific to a childhood spent with alcoholic parents; it may spring from diverse dysfunc-
tional family environments. Indeed, Morgan (1991) has identified the causes of
codependence as “earlier life experiences that frequently involved childhood abuse of
some type” (p. 721). The proposed link between codependence and childhood abuse
nevertheless lacks adequate empirical documentation. A recent study by Fischer,
Wampler, Lyness, and Thomas (1992) attempted to address this issue, but found that
codependence was not predicted significantly by dysfunctional patterns in the family
of origin. The present study sought to extend this line of enquiry with specific attention
to childhood trauma. The following hypothesis was formulated.

HYPOTHESIS
1

Codependence in adulthood is predicted by a childhood with an alcoholic parent

and, more generally, by a diversity of childhood traumatic experiences.
A second issue for the study concerned the relationship between codependence and
a more widely recognized personality disorder. One of the criticisms frequently leveled
at the concept of codependence is that its proponents do not make clear whether
codependence is best regarded as a psychological disorder, as a personality trait, or as
a social condition (Gierymski & Williams, 1986; Gomberg, 1989; Haaken, 1990; Harper
& Capdevila, 1990). Certainly some endeavors to operationalize codependence (Fischer,
Spann, & Crawford, 1991; O’Brien & Gaborit, 1992) suggest that codependent charac-
teristics are trait-like and exist to varying degrees across the whole population. At the
same time, Cermak (1991a) argues that codependence is both a personality trait and
a personality disorder. This situation is consistent with the approach of the DSM
(American Psychiatric Association, 1987), whereby personality traits may become so
rigid and maladapative that they represent personality disorders. By way of illustra-
tion, narcissistic traits are evident in varying degrees across the whole population, and
if these narcissistic tendencies significantly impair a person’s life a diagnosis of narcissistic
personality disorder may be applicable. In much the same spirit Cermak (1986), while
recognizing codependent characteristics as traits, has constructed DSM-like criteria for
a proposed diagnostic category of codependent personality disorder.
In addition, Cermak (1991a, 1991b) has posited a relationship between codependence
and narcissism. Specifically, codependent traits are held to be the complement of nar-
cissism. According to Cermak, both codependence and narcissism arise during the young
child’s symbiotic phase of development and serve to block progression to the separation-
individuation phase. Both sets of traits involve defective “mirroring”: narcissistic people
relate to others only by seeing aspects of themselves mirrored in these people, whereas
codependent people seek to relate to others by mirroring them. The origins of
codependence and narcissism also involve defective mirroring, in this instance by parents.
If parents deny their child’s capabilities (the need to be appreciated) and reinforce the
child’s value only as a reflection of themselves (the need to be appreciative), the child
will develop codependent tendencies. On the other hand, if the parents validate the child’s
capabilities but deny the child‘s idealized image of the parents, the child will develop
a narcissistic personality structure. Although Cermak believes that every person suffers
both codependent and narcissistic L ‘ ~ ~ ~ to n dsome
~ ’ ’degree, the underlying processes
are inherently complementary, and, thus, codependence and narcissism are inversely
related.
660 Journal of Clinical Psychology, September 1995, Vol. 51. No. 5

Cermak’s (1991a, 1991b) theory of the development of codependent and narcissistic

traits formed the basis for the study’s second hypothesis, which may be formulated as
f 0110ws .

HYPOTHESIS
2

Narcissistic traits are negatively weighted predictors of codependent tendencies.

METHOD
Subjects
The study involved the face-to-facc administration of an inventory of question-
naires. The participants were recruited as a convenience sample of Australian adults.
The sample comprised 100 women and 90 men, who ranged in age from 18 to 70 years
(M = 31.5, median = 28.5); a significant minority were university students.
Materials
The test inventory comprised seven questionnaires. One was a brief form asking
for details of gender and age. Two questionnaires were self-report measures of
codependence, two others were measures of narcissism, another questionnaire surveyed
childhood trauma associated with living in the household of an alcoholic parent, and
the remaining questionnaire assessed the incidence of a wide range of other childhood
trauma.
The quantitative assessment of codependent traits is still in an early stage of develop-
ment; no single scale has emerged as the standard index of codependence. For this reason
two measures of codependent traits, the CDI and the SFCDS, were in the study. The
CDI or Codependency Inventory was developed by O’Brien and Gaborit (1992) and com-
prises 17 dichotomous (TrueIFalse) items that focus on the appropriateness of interper-
sonal relationships and self-autonomy. This scale is reported to have satisfactory inter-
nal consistency reliability (Cronbach’s alpha = .@; O’Brien & Gaborit, 1992). The
SFCDS or Spann-Fischer Codependency Scale (Fischer et al., 1991) has 16 items with
responses to be made on a 6-point Likert scale (1 = “Strongly disagree” to 6 = “Strongly
agree”). The internal consistency reliability of the SFCDS is adequate (Cronbach’s alpha
= .73 to .80), and some empirical evaluation suggests that the scale has convergent and
discriminant validity (Fischer et al., 1991). The higher the score on the CDI or the SFCDS,
the stronger are the respondent’s codependent traits.
The most widely used measure of narcissistic traits (as opposed to a diagnostic in-
strument for narcissistic personality disorder) is the Narcissistic Personality Inventory
or NPI (Raskin & Terry, 1988). In its most recent revision the NPI comprises 40 pairs
of statements; from each pair respondents must select the statement with which they
most agree. High scores indicate strong narcissistic tendencies. The internal consistency
reliability of the NPI is satisfactory (Cronbach’s alpha = .83), and the scale’s convergent
and construct validity has been empirically substantiated with clinical and nonclinical
samples (Emmons, 1987; Raskin & Terry, 1988; Watson, Grisham, Trotter, & Bider-
man, 1984).
A second index of narcissism, the Narcissistic Personality Disorder scale or NPD,
was used in the study because this measure was designed more specifically for the
diagnosis of narcissistic personality disorder and yet is reported to be uncorrelated with
the NPI (Chatham, Tibbals, & Harrington, 1993; Emmons, 1987; Mullins & Kopelman,
1988; Watson et al., 1984). The NPD was developed by Ashby, Lee, and Duke (1979)
and comprises 19 items from the MMPI that were found by these researchers to
discriminate narcissistic personality disordered patients from other diagnostic groups.
The internal consistency of the NPD is reported to be rather variable (KR-20 = .42
 Codependence, Narcissism, and Childhood Trauma 66 1

to .81), but the questionnaire has been validated with clinical and nonclinical samples
(Ashby, Lee, & Duke, 1979; Mullins & Kopelman, 1988; Solomon, 1982). High scores
on the NPD are held to be indicative of a narcissistic personality disorder.
The Children of Alcoholics Screening Test or CAST (Jones, 1983) represents an
index of the extent of childhood trauma associated specifically with a parent’s alcohol
use. The CAST has 30 dichotomous (yedno) items that address the respondent’s
childhood feelings, behavior, and experiences that arose in the context of living with
a parent who had a drinking problem. High scores indicate that the parent’s alcohol
use had a substantial impact on the respondent during childhood.
Childhood truma in a more general context were indexed by the Survey of Traumatic
Childhood Events or STCE (Council & Edwards, 1987). The 30 items of the STCE tap
childhood trauma of 1 1 factorially determined types: intrafamilial sexual abuse, ex-
trafamilial sexual abuse, intrafamilial physical abuse, loss related to a friend, loss related
to the family, isolation, personal illness or accident, parental divorce/separation and
abortion/miscarriage, assault, loss of the home, and robbery. Responses are made on
a 5-point Likert scale (1 = none; 5 = more than ten). For each of the 11 subscales
a score is computed by summation across items; high scores attest to a high incidence
of the given class of childhood trauma. Psychometric data for the STCE currently are
unavailable, but the STCE is the most broadly based and professionally constructed
measure of childhood trauma yet developed.
Procedure
Participants were tested individually in a relatively quiet location, usually the person’s
residence or workplace. A “plain language” informed consent form was attached to the
front of the inventory; this sheet explained the objective of the study, drew attention
to the personal nature of items in the STCE and the CAST, and stressed that participa-
tion was voluntary and confidential. An appeal was made to participants to respond
as spontaneously and openly as possible. For each questionnaire in turn the procedure
was explained, and the participant then completed that section of the inventory.
On completion of the test inventory, participants were offered immediate
psychological support for any issues that arose from the study.

RESULTS
Descriptive statistics (mean and standard deviation) for each of the research measures
are presented in Table 1.
Before we address the data in terms of the two hypotheses it is appropriate to report
the relationship between the two measures of codependent traits. The Pearson correla-
tion coefficient between the CDI and the SFCDS was .63. While this coefficient differs
significantly from zero (p < .OOOOOl), it is not large, particularly if the CDI and SFCDS
are taken to be tapping the same dimension of psychological functioning. Thus, the
CDI and the SFCDS have barely 40% in common variance. This result reaffirms the
point that any conclusion on the nature of codependence is best founded upon empirical
investigation with a variety of indices of codependent traits.
Hypothesis 1 implies that scores on the codependence scales may be predicted by
those on the surveys of childhood trauma and the experience of a parent’s alcohol use.
A standard multiple regression was performed between the CDI as the dependent variable
and the CAST, the 1 1 STCE scales, gender, and age as independent variables. In a stand-
ard multiple regression all independent variables are entered into the regression equation
simultaneously; this is the recommended method when there are insufficient theoretical
grounds for controlling the order of entry of variables (Tabachnick & Fidell, 1989).
Analysis was conducted using SYSTAT (1992) MGLH/Regression software. Tolerance
statistics for the analysis were acceptable and ranged from .62 to .90;multicollinearity
662 Journal of Clinical Psychology, September 1995, Vol. SI, N o . 5

Table 1
Descriptive Statistics f o r Each Experimental Variable (N = 190)

Variable M S

CDI 6.82 3.83

SFCDS 52.21 13.84
NPI 1 1 :74 6.34
NPD 7.26 3.42
CAST 6.41 8.75
STCE I : Intrafamilial sexual abuse 3.84 2.34
STCE 2: lntrafamilial physical abuse 16.68 6.53
STCE 3: Loss related to friend 3.05 I .40
STCE 4: Extrafamilial sexual abuse 4.41 2.40
STCE 5: Loss related to family 5.37 1.83
STCE 6: Isolation 3 .I7 5.94
STCE 7: Personal illness or accident 3.03 1.26
STCE 8: Parental divorce or separation; abortion or miscarriage 2.42 .74
STCE 9: Assault 2.90 I .22
STCE 10: Loss of home 4.60 1.39
STCE 1 1 : Robbery I .09 .3I

of predictor variables, therefore, was of no practical concern (Darlington, 1990). Never-

theless, only 6% (or 0% adjusted) of the variability in CDI performance was predicted
by scores on the independent variables. The multiple correlation R for the regression
did not differ significantly from zero, F(14,175) = .76, p = .71. A similar regression
analysis with the SFCDS as the dependent variable yielded essentially the same outcome,
RZ = .07; adjusted RZ = .OO;F(14,175) = .95, p = .51.
Given that the concept of codependence evolved in the context of the families of
alcoholics, the failure of the CAST to predict codependence scores is remarkable. Closer
inspection of the CAST data indicated a substantial proportion of participants who had
a score of zero on this test; the distribution of CAST scores, therefore, was positively
skewed to a substantial degree. To assess this feature’s impact on the results of the regres-
sion analyses, the CAST scores were recorded to produce two groups of participants,
those who reported at least some negative experience of a parent’s alcohol use and those
who did not. Post hoc standard multiple regressions with the recoded CAST data still
produced nonsignificant results. Thcrefore, the lack of support for hypothesis 1 is not
a mere artifact of the skewness of the CAST data.
Under hypothesis 2, scores on the codependence scales should be predictable by
(but negatively related to) those on the two measures of narcissism. Regression analysis
again was used as the statistical technique for testing the hypothesis. In this instance,
however, two of the independent variables NPI and NPD supposedly relate to the same
dimension, and the potential problem of multicollinearity arises. In a preliminary analysis
the Pearson correlation between the NPI and the NPD was computed; this coefficient
was - .06 @ = .39). The finding of no statistical relationship between two indices of
narcissism is notable and will be addressed later, but for present purposes it is sufficient
to observe that collinearity between the narcissism indices does not emerge as an im-
pediment to the analyses for hypothesis 2.
A standard multiple regression was performed between the CDI as the dependent
variable and the NPI, the NPD, gender, and age as independent variables. Table 2
 Codependence, Narcissism, and Childhood Trauma 663

presents the unstandardized regression coefficients (B) and intercept, the standardized
regression coefficients (p), the semipartial correlations (s?), and R, R2 and adjusted R2;
a semipartial correlation represents the contribution of an independent variable to R2
when the contribution of the other independent variables is taken out of both the depend-
ent variable and the particular independent variable. The tolerance statistics for this
analysis were satisfactory and ranged from .90 to .98; multicollinearity of independent
variables, therefore, is not a problem. The multiple correlation R for the regression was
significantly different from zero, F(4,185) = 8.71, p -c .000005. Two of the independ-
ent variables contributed significantly to the prediction of CDI scores. The effective
predictors were NPI (s? = .020) and the NPD (sr' = .120). In total, 16% (or 14%
adjusted) of the variability in CDI performance was predicted by scores on the nar-
cissism scales, gender, and age.

Table 2
Standard Multiple Regression of Narcissism (NPI, NPD), Gender, and Age on the CDI Index
of Codependency (N = 190)

Variables B P P Sr2
(unique)

NPI - .089 - .15 .039 .020

NPD ,393 .35 .OOoOl .120
Gender ,183 .02 .734
Age - .030 - .08 .224
Intercept = 5.676

R2 = .16. Adjusted R2 = .14. R = .40,p = .000002.a

aReported p value is for a two-tailed test of the hypothesis that R = 0.

A similar regression analysis was conducted with the SFCDS as the dependent
variable. The results of this analysis are summarized in Table 3. The multiple correla-
tion R for the regression was significantly different from zero, F(4,185) = 27.26,
p < .OOOOOl. In this case only one of the narcissism measures, the NPD, contributed
significantly to the prediction of SFCDS scores (s? = .337), although the contribution
by the NPI was of borderline significance (s? = .012). Altogether, 37% (or 36% ad-
justed) of the variability in SFCDS scores was predicted by the narcissism scales, gender,
and age.

Table 3
Standard Multiple Regression of Narcissism (NPI, NPD), Gender, and Age on the SFCDS Index
of Codependency (N = 190)

Variables B P P S?
(unique)

NPI - .257 - .12 .057 .012

NPD 2.316 .59 .ooOol .331
Gender - 2.069 - .07 .218
Age .031 .02 .684
Intercept = 40.045

R2 = .37. Adjusted RZ = .36. R = .61, p = .OOOOO1.a

aReported p value is for a two-tailed test of the hypothesis that R = 0.

664 Journal of Clinical Psychology, September 1995, Vol. 51, No. 5

The data of Tables 2 and 3 nevertheless offer contradictory support for hypothesis
2. As evidenced by the signs of the regression coefficients, in both analyses codependence
scores were related negatively to the NPI, but related positively to the NPD. Under
hypothesis 2 these relationships were predicted to be uniformly negative.

DISCUSSION
Several results from the statistical analysis are noteworthy. The correlation between
the two measures of codependence (.63), despite its statistical significance, is not impres-
sive, and it certainly signals the need for further work on the measurement of codependent
tendencies. At the same time the consistency of results for the two indices of codependence
provides reasonable grounds for drawing some strong conclusions from the study.
The lack of support for hypothesis 1 is problematic for the theoretical framework
of codependence. None of the incidences of a wide variety of childhood trauma, nor
the level of impact of living with an alcoholic parent, contributed to the prediction of
codependent traits in adulthood. This finding is consistent with the report by Fischer
et al. (1992) that codependence had no statistical relationship with dysfunctional patterns
in the family or origin. Of course, these results cannot be taken to indicate that childhood
trauma are inconsequential for adult adjustment, but they do cast doubt on the posited
link between childhood trauma and codependent tendencies in adulthood. Further
research is required to ascertain whether codependence has other origins or, indeed,
whether codependence is coherent as an explanatory psychological concept.
The nonsignificant ( - .06) correlation between NPI and NPD scores, in conjunc-
tion with several other reports of the lack of a relationship between these two measures
of narcissism (Chatham et al., 1993; Emmons, 1987; Mullins & Kopelman, 1988; Watson
et al., 1984), may be taken to suggest that narcissism is a complex, probably multifaceted,
personality domain. The multidimensionality of narcissism is promoted explicitly by
Raskin and Terry (1988) and is a consideration in the interpretation of the results of
analyses for hypothesis 2. It will be recalled that the data here were reasonably consis-
tent in confirming the contribution of narcissism to the prediction of codependence,
but only the NPI scores carried the negative weighting implied by Cermak’s (1991a, 1991b)
theory. Contrary to the prediction under hypothesis 2, NPD scores yielded positive regres-
sion coefficients. Not only are the NPD data contrary to Cermak’s account, but they
proved to be the most powerful of the predictors in the two analyses. (See Tables 2 and
3.) In addition, the proportion of the variance in codependencc scores for which nar-
cissism accounted is not as high as might have been expected under Cermak’s model.
Although it appears that there is now some empirical basis for positing a link between
codependence and narcissism, there remains a need for further research designed to deter-
mine which specific facets of narcissism are related to codependence and what distinct
developmental processes underlie these links.
The findings of the study seem to weaken substantially some of the current theoretical
underpinnings of the concept of codependence. This has implications for the rather in-
discriminate application of the concept in current mental hezlth practice. Certainly
“codependence” may well have heuristic value in labelling the pain experienced by people
in dysfunctional relationships and in providing a readily comprehensible conceptual con-
text for mutual support activities, but until a sound empirical base is established any
extensive reliance upon the concept of codependence for psychotherapeutic purposes
should be considered with caution.

REFERENCES
AMERICANPSYCHIATRIC ASSOCIATION (1 987). Diagnostic and statistical manual for mental disorders (3rd
ed. rev.). Washington: Author.
 Codependence, Narcissism, and Childhood Trauma 665

ASHBY,H. U, LEE, R. R., & DUKE,E. H. (1979, September). A narcissistic personality disorder MMPI
scale. Paper presented at the annual meeting of the American Psychological Association, New York.
CERMAK, T. L . (1986). Diagnostic criteria for codependency. Journal of Psychoactive Drugs, 18, 15-20.
CERMAK, T. L. (1991a). Co-addiction as a disease. Psychiatric Annals, 21, 266-272.
CERMAK,T. (1991b). The relationship between codependence and narcissism. In T. M. Rivinus (Ed.), Children
of chemically dependent parents: Multiperspectivesfrom the cutting edge (pp. 131-152). New York:
Brunner/Mazel.
CHATHAM, P. M., TIBBALS, C. J., & HARRINGTON, M. E. (1993). The MMPI and the MCMI in the evalua-
tion of narcissism in a clinical sample. Journal of Personality Assessment, 60, 239-25 1.
COUNCIL, J. R., dr EDWARDS, P . W. (1987). Survey of Truumatic ChildhoodEvents. Unpublished measure.
Fargo: North Dakota State University.
DARLINGTON, R. B. (1990). Regression and linear models. New York: McGraw-Hill.
EMMONS, R. A. (1987). Narcissism: Theory and measurement. Journal of Personality and Social Psychology,
52, 11-17.
FISCHER,J. L., SPA”, L., & CRAWFORD, D. (1991). Measuring codependency. Alcoholism Treatment
Quarterly, 8, 87-100.
FISCHER,J. L., WAMPLER, R., LYNESS,K., &THOMAS, E. M. (1992). Offspring codependency: Blocking the
impact of the family of origin. Family Dynamics of Addiction Quarterly, 2, 20-32.
GIERYMSKI, T., a WILLIAMS, T. (1986). Codependency. Journal of Psychoactive Drugs, 18, 7-13.
GOMBERG, E. S. L. (1989). On terms used and abused: The concept of “codependency.” Drugs and Society,
3(3/4), 113-132.
HAAKEN, J. (1990). A critical analysis of the co-dependence construct. Psychiatry, 53, 396-406.
HAAKEN, J. (1993). From Al-Anon to ACOA: Codependence and the reconstruction of caregiving. Signs:
Journal of Women in Culture and Society, 18, 321-354.
HARPER,J., & CAPDEVILA, C. (1990). Codependency: A critique. Journalof Psychoactive Drugs, 22,285-292.
HORNEY,K. (1942). Self-analysis. New York: Norton Press.
JONES,J. W. (1983). Children of Alcoholics Screening Test: Test manual. Chicago: Camelot Unlimited.
KRESTAN, J., & BEPKO,C. (1990). Codependency: The social reconstruction of female experience. Smith
College Studies in Social Science, 60, 2 16-232.
MORGAN, J. P . (1991). What is codependency? Journal of Clinical Psychology, 47, 720-729.
MULLINS, L. S., & KOPELMAN, R. E. (1988). Toward an assessment of the construct validity of four measures
of narcissism. Journal of Personality Assessment, 52, 610-625.
MYER,R. A., PETERSON, S. E., & STOFFEL-ROSALES, M. (1991). Co-dependency: An examination of underlying
assumptions. Journal of Mental Health Counseling, 13, 449-458.
O’BRIEN,P. E., & GABORIT, M. (1992). Codependency: A disorder separate from chemical dependency. Journal
of Clinical Psychology, 48, 129-136.
PREST,L. A., 81 PROTINSKY, H. (1993). Family systems theory: A unifying framework for codependence.
American Journal of Family Therapy, 21, 352-360.
RASKIN,R., & TERRY,H. (1988). A principal-components analysis of the Narcissistic Personality Inventory
and further evidence of its construct validity. Journal of Personality and Social Psychology, 54,890-902.
SCHAEF,A. W. (1986). Co-dependence: Misunderstood, mistreated. San Francisco: Harper & Row.
SOLOMON, R. S. (1982). Validity of the MMPI Narcissistic Personality Disorder scale. Psychorogical Reports,
50, 463-466.
SYSTAT(1992). Statistics, version 5.2 edition. Evanston, IL: SYSTAT.
TABACHNICK, B. G., & FIDELL,L. S. (1989). Using multivariate statistics (2nd ed.). New York: Harper & Row.
WATSON, P. J., GRISHAM, S. O., TROTTER, M. V., & BIDERMAN, M. D. (1984). Narcissism and empathy:
Validity evidence for the Narcissistic Personality Inventory. Journal of Personality Assessment, 48,301-305.