Psychological Reports, 1992, 70, 839-849.

O Psychological Reports 1992

PERSONALITY DISORDER CONCEPT SCALES AND 16 PF DIMENSIONS ' JAMES E. DELAMATRE AND JAMES M. SCHUERGER Ckvekand State UniversiQ Summary.-Using 148 clients from a private practice sample, concurrent scores on the Morey, Waugh, and Blashfield Minnesota Multiphasic Personality Inventory Scales and the Sixteen Personality Factor Questionnaire were factor analyzed in an attempt to map the Axis I1 disorders within a circumplex o Interpersonal behavior. Individual predictions about the 16 PF primaries were made and then equations were developed to predict the Morey, Waugh, Blashfield MMPI scales from the 16 PF scores. Using a separate sample of 43 clients and 21 students, a uoss-validation of concepts was performed by using the 16 PF to generate the MMPI scores and correlating the predicted scores with concurrent MiUon Clinical Multiaxial Inventory scores. Concordance of the personality disorder characterizations across the instruments was high. Many of the predicted personality disorder characteristics were supported by the research, and 10 of 11 correlations with the Millon were significant. This research suggests the 16 PF is useful for generating hypotheses about disorders and that the Axis I1 conceptualizations fit as expected within normal personality space.

f'

Several authors have noted that the Sixteen Personality Factor Questionnaire (16 PF) is useful for generating hypotheses about the diagnosis of personality disorders (Meyer, 1989; Schuerger, Watterson, & Croom, 1990; Fowler, 1989). The purpose of this research was to develop a conceptual map of the I1 disorders (DSM-111-R; American Psychiatric AssociaI tion, 1987) in terms of the 16 PF factors. For this research, the Axis I1 concepts were defined by the Morey, Waugh, and Blashfield (MWB) Minnesota Multiphasic Personality Inventory (MMPI) Scales (Morey, Waugh, & Blashfield, 1985), and the personality disorder scales from the Millon Clinical Multiaxial Inventory. A further was to extend prior research on the concordances between scores on the MWB MMPI scales and on the MilIon inventory (McCann, 1989, 1991; Zarrella, Schuerger, & Ritz, 1990). Meyer (1989) provides a rational model for clinical interpretation of the 16 PF factors which would be consistent with persondty disorder constructs. Other authors (Schuerger, et al., 1990; Fowler, 1989) have further developed designs which utilize the 16 PF for generating hypotheses about personality disorders. There is considerable agreement among these authors about 16 PF attributes which characterize the various disorders, but there are some inconsistencies in point of view and no over-arching conception of the area. The aim of this study was to advance the conceptual map linking

xis

'Address correspondence to James M. Schuerger, Cleveland State University, Psychology Department, 215D Stdwell Hall, Cleveland, Ohio 44115.

840

J. E. DELAMATRE & J. M. SCHUERGER

the 16 PF normal domain to the more extreme domain of the ~ e r s o n a l i t ~ disorders. MODEL Development of a model for this research was guided by concerns for (1) conceptual generality and (2) clinical utility. Relevant to these concerns is the fact that the 16 PF provides scores at two levels, that of the 16 primary factors and that of the five secondary factors derived as linear composites of the primaries. For practitioners familiar with the 16 PF, a conceptual map relating the 16 PF domain to the disorders at either level would have utility. For others, the more general secondary factors might be expected to have greater utility. Let us be more specific about the latter case. The 16 PF secondary factors are readily linked with various representatives of the "Big Five" personality dimensions on which researchers are now converging (Digman, 1990; Costa & McCrae, 1990; Schuerger & Allen, 1986). Furthermore, recent work which related these broad dimensions to Axis I1 scales (Pincus & Wiggins, 1990; Wiggins & Pincus, 1989) enabled us to make an explicit model of expected relationships between 16 PF broad factors and the Personality Disorder scales. The model and correlative results for this study are presented in Table 1, along with results from an admirable study by Wiggins and Pincus (1989). These authors factor analyzed the NEO Personality Inventory (Costa & McCrae, 1990) to represent the "Big Five" and the MWB MMPI scales. Their results as factor loadings are in columns labeled W under each of the 16 PF broad factors. N E O Personality Inventory names for the putatively similar variables are also provided at the bottom. Wiggins and Pincus' (1989) results for two NEO Personality Inventory factors of Openness and Agreeableness were reflected to align the directionality of factors with the data of the present study. This model has been linked to an even more general model developed years ago by Leary and more recently by Wiggins (both cited by Wiggins and Pincus, 1989) in which the personality disorders are presented in terms of a circumplex of interpersonal behaviors and attitudes. The circumplex is delineated by the horizontal dimension of cold-warm and the vertical dimension of submissive-dominant. The basic circumplex model is very general and is seen in a wide variety of data (Keisler, cited by Pincus & Wiggins, 1990; McLemore & Brokaw, 1987; Mdon, cited by Wiggins & Pincus, 1989; Pincus & Wiggins, 1990). Fig. 1 presents the interpersonal circumplex with personahty hsorders projected onto it. From the Wiggins and Pincus study, the disorder acronyms forming our model begin with w, the present study's MWB MMPI concepts begin with m, and the 16 PF broad factors are boxed. At the level of the primary 16 PF variables, a model of the Axis I1 disorders was developed for the study drawing on the work of Fowler (1989),

TABLE 1 VARIMAXROTATEDFACTOR LOADINGS FORMMPI, 16 PF, AND MILLONINVENTORIES Study

I (EXV) MP MC

W 16 PF Secondary Factors Extroversion Anxiety Practicality Independence Self-control MWB MMPI Scales Narcissistic Histrionic Borderline Antisocial Dependent Obsessive-Compulsive Passive-Aggressive Paranoid Schizotypal Avoidant Schizoid

NEO Personality Inventory

80

80 -36

11 (ANX) MP MC

W

111 (PRAC) MC

W* MP

90

70

68

80 93

67 80 35 -39

-59 -80

64

35

-82

58

-78 -70 -86

Extraversion

63 37

61

86 74 64 82 81 84 58 52

68

43

-49 72

-76 -77 -93

V (CONT) W

MP

88

92

-41

-26

MC

-35 96

-35 50 81

IV (IND) MP MC

80 -80

50

W*

33 65 -57

39 37 -38

75 75 -60 -81 89

Neuroticism

85

-60 -30

46

40

-43 51

Openness

Agreeableness

Conscientiousness

Note.-W = W~gglnsloadings, MP = 16 PF and MWB MMPI loadings, MC = 16 PF and Millon loadings. The Wiggins loadings marked with an asterisk are reflecretl

J. E.DELAMATRE & J. M. SCHUERGER

ASSURED. DOMINANT

INTROVERTED

UNASSURED SUBMISSIVE

FIG.1. W P i n s ' ~ r o j e c t i o n sof the Minnesota Multiphasic Personality Inventory (MMPI) personality disor er sc es (beginning with "w"),. current study using the Morey, Waugh, and Blashfield MMPI scales (beginning with "m"), and current study of 16 PF Secondary Factors (boxed)

Meyer (1989), and Schuerger, et al. (1990). Where there was disagreement among the authors about the relevant factor and the direction of the relationship, a democratic or intuitive decision prevailed. For example, Meyer (1989) states that low N and Meyer (1989) and Schuerger, et al. (1990) that low Q, would indicate denial of tension in paranoid and compulsive personalities, but the authors decided high N and Q, might show the vigilance characteristic of these disorders. For antisocial and schizotypal personality disorders, Meyer (1989) indicates A would be higher, but since no other authors considered this variable it was dropped from the predictions. For borderline personahty disorder, Q, was similarly dropped. The relationships and directions expected are indicated by a plus or minus sign above the obtained correlations in Table 3 below.

METHOD Sample The initial sample consisted of 148 clients from several suburban private practices who had concurrently taken MMPI and 16 PF questionnaires sometime during the course of therapy. The inventories were administered between January 1, 1988 and October 30, 1990. There were 79 men and 69 women, and the over-all age range of the clients was from 15 to 76 years old (M = 36, SD = 11). A separate sample was used to cross-validate the findings. This sample consisted of 43 patients from one suburban private practice and 21 undergraduate students who had concurrently taken the 16 PF and the Millon in-

PERSONALITY DISORDERS AND THE 16 PF

843

ventory. These inventories were administered between July 25, 1986 and March 4, 1991. There were 30 men and 35 women in this sample, and the age range of the private practice portion was 19 to 91 years (M = 43, SD = 16). The ages were not available for the students in this sample.

Instruments The 16 Personality Factor Questionnaire (Cattell, Eber, & Tatsuoka, 1970) is a normal domain personality inventory derived by selecting a set of personality trait descriptors, intercorrelating behavioral ratings and questionna& answers, and factor analyzing the final set. Sixteen primary factors were named according to the factor content and its correlates. The 16 primary factors also combine into five broad factors which are analogous to the "Big Five" personality dimensions. The MMPI Personality Disorder Scales are 11 scales derived by a rational/empirical approach based on the DSM-I11 Axis I1 descriptions (Morey, et al., 1985). Overlapping and nonoverlapping sets have been used in research, but both perform similarly and there is no consensus on the utility of either set. This study used the overlapping scales. For a psychiatric inpatient sample, Hurt, Clarkin, and Morey (1990) found the temporal stability of the overlapping scales was similar to that of the MMPI clinical scales. In a preliminary validity study by Morey, Blashfield, Webb, and Jewell (1988) scores on the overlapping Borderhe, Antisocial, and Dependent MWB scales discriminated effectively between normals and other personality disorders, scores on the Histrionic scale discriminated from other personality disorders but not from normal scorers, and the Passive-Aggressive scale discriminated from the normal group but not from other personality disordered groups. McCann (1991) factor analyzed the overlapping and nonoverlapping MWB and MCMI-I1 scales for a psychiatric inpatient sample and reported high convergent and discriminant validity on nine of the MWB scales. The exceptions were the Compulsive scale which did not possess convergent or discriminant validity and the Paranoid scale which had low discriminant validity. Item overlap had an individual effect on scale valilty, but no general effect. The Millon Clinical Multiaxial Inventory (Millon, 1982) requires self-report on clinical domain and is based on Millon's personality typology. The concepts are similar to DSM-111-R descriptions, but there is some disagreement about the nature of the relationship (Morey & Levine, 1988). The Millon inventory shows internal consistency and temporal stability similar to those of other self-report questionnaires (Millon, 1982). Data from four previous studies have been consistent in showing substantial correlations of scores on the MWB MMPI scales and on the MiIlon scales (McCann, 1989, 1991; Streiner & Miller, 1988; Zarrella, et al., 1990). Exceptions are the Compulsivity scales, on which scores usually correlate negatively across the

J. E. DELAMATRE & J. M. SCHUERGER

844

instruments, and the Antisocial, Borderline, and Paranoid scales, on which scores correlate positively but frequently below acceptable statistical significance. The data are summarized in Table 2. The Pearson correlations are averaged across the four studies and are at about the level of the test-retest reliability for comparable scales after an interval of six months (Schuerger, ZarrelIa, & Hotz, 1989; Schuerger, et a/., 1991). One concern of the present study was whether optimum weighting of 16 PF primaries could approximate these results. TABLE 2 MILLONINVENTORY AND MMPI PERSONALITY DISORDER SCALES: PEARSON CORRELA~ONS AVERAGEDACROSSSTUDIES

Scale Narcissistic Histrionic Borderline Antisocial Dependent Com~ulsive

r

.70 .69 .54 .29 .55 -.30

Scale

Passive-Aggressive Paranoid Schizotypal Avoidant

Schizoid

r

.62 .34 .64 .77 .66

Procedure Patients' MWB MMPI and 16 PF scores were matched and coded into a single file containing only the T scores to protect the clients' anonymity. Following the model of Wiggins, the 16 PF second-order scores and the MWB Axis I1 concepts were factor analyzed and rotated to the varimax criterion, using default procedures in SPSS/PC. This procedure was repeated with the 16 PF and Millon scores. To project the MWB MMPI scores onto the interpersonal circumplex, the two factors with highest loadings on 16 PF Extroversion (EXT) and Independence (IND) were rotated again to give Independence maximum loading on the vertical (Dominance) axis. This procedure was also repeated with the Millon scores. The x and y coordmates for variables in Fig. 1 above are the resultant loadings for this study and the Wiggins and Pincus study (1989) for this model. For relating the 16 PF primaries to Axis I1 concepts, all 16 primaries and 11 MWB MMPI scales were correlated using Pearson product-moment procedure. Given the disagreement in the literature about the direction of the relationship of some variables and the number of correlations involved, a nondirectional alpha level of .O1 was set. In addition, a forward stepwise multiple regression [SPSS/PC] was performed using each of the 11 MWB MMPI scales as the dependent variables. Predictors consisted of 16 PF primary factors, singly for some primaries and in efficient, synthetic clusters for others. The clusters were created from the authors' experience with the instrument in similar predictive settings and are based on correlations among

845

PERSONALITY DISORDERS AND T H E 16 PF

the primaries. Clustering in this fashion eliminates the possibility of suppressor variables, which are difficult to interpret, and minimizes mdticolinearity. The synthesized variables are similar to the related 16 PF second-order factors. The cross-validation sample was then used to generate predicted MWB scores and correlate them with the Millon inventory scores.

RESULTS The correlation matrix is presented in Table 3. Many of the hypotheses were supported, although two of 16 PF variables correlated in the opposite direction than was predicted rationally. For eight of 11 scales, at least four of the correlations were significant in the predicted direction: Narcissistic, Histrionic, Antisocial, Dependent, Borderline, Schizotypal, Avoidant, and Schizoid. At least two of the directional predictions were significant for all of the scales except Paranoid. Note that none of the predictions about a scale were borne out entirely and that several were significant in the direction opposite to those expected. The most notable exception to the expected results concerns the Histrionic scale in that the 16 PF Factors C and Q, correlated the opposite direction expected. Other exceptions were Factor H on the Passive-Aggressive and Paranoid scales, Q, on Schizotypal, and C on Schizoid scales. TABLE 3

EXPECTED AND OBTAINED CORRELA~ONS 16 PF Morey, Waugh, and Blashfield MMPI Scales Factors NAR HST BDL ANT DEP CPS PAG PAR STY AVD SZD

M

SD

(continued on next oaee) Note.-NAR: Narcissistic, HST: Histrionic, RDL: Borderline, ANT: Antisocial, DEP: Dependent, CPS: Compulsive, PAG: Passive.Aggressive, PAR: Paranoid, STY: Schizotypal, AVD: Avoidant, SZD: Schizoid. Expected correlnr rons are indicated by the directional sign immediately above the value. *p