Competing Models Of Personality Disorders

This paper reviews the virtues of applying a dimensional model and/or the more traditional categorical model to classification of personality disorders. The authors suggest how the particular merits of both might be incorporated. A model is proposed that organizes personality typology around a dimension of severity of impairment or dysfunction. This model would utilize the categorical model for those patients with the more severe functional failures for which traditional etiological research is most likely to bear fruit, and it would utilize the dimensional model for less functionally impaired patients whose psychopathology merges with normalcy

A visible and articulate group of colleagues (Clarkin et al., 1983: Frances, 1982; Frances & Cooper, 1981; Frances & Widiger, 1986a, b; Widiger & Frances, 1985) have suggested that personality disorders are essentially continuous with each other and that a dimensional model is best suited for conceptualizing their taxonomy. Frances and Widiger ( 1986b), in fact, have confidently predicted "that dimensional approaches will gradually supplant the categorical in the classification of personality disorders" (1986, p. 396). As such these authors are entering a debate that has a long, colorful history and are joining other prominent exponents of the dimensional model (Eysenck, Wakefield, & Friedman, 1983; Hine & Williams, 1975; Kendall, 1975; Mezzich, 1979; Strauss, 1973). A good review of this literature goes beyond the purpose of this paper. Here we wish to redress the balance by reexamining the arguments for a dimensional model and comparing its virtues and limits to those of the more traditional categorical model.

We believe the merits of these two models of personality disorder should be measured by whether one model communicates more information about cause, pathogenesis, course, prognosis, and treatment than the other. As such this review judges these models on the basis of their claims or aptitude for external validation and clinical utility. After examining the relative strengths of the dimensional and categorical approaches, a synthesis of the two models is proposed that we believe can further the development of psychiatric classifications for personality disorders.

THE DIMENSIONAL MODEL

This model sees the types of personality as having intrinsically gray boundaries and the effort to divide them into discrete categories as inherently arbitrary and distorting to their true nature. Figure 1 illustrates how three of the existing DSM-III-R (American Psychiatric Association, 1987) categories could be seen to exist on a continuum. In the dimensionalist's view the separation of borderline, histrionic, and dependent into these categories may helpfully communicate quantitative differences and hence have practical value, but this separation also conveys a misleading sense of distinctiveness. Widiger and Frances (1985) have proposed that reliance upon the dimensional model might resolve many of the current problems with personality disorder diagnosis. An initial step in this direction, suggested by Stone (1980) and by Millon (1981), could be to "dimensionalize" the current list of 1 1 DSM-III-R personality types. Thereby, a patient could be rated on the extent to which he or she has each of these maladaptive personality traits.

The suggested advantages of the dimensional model are the following:

A. Dimensional subtyping lends itself to numerical representation and has a continuous distribution.The dimensional model accepts the enorrnous diversity of personality traits and the great range of possible com binations amongst them found in nature. As such, this model may do more justice to the uniqueness of individuals. Dimensional models have enjoyed great popularity within academic psychology since T. Leary's interpersonal circumplex (1957). By examining a broad range of personality traits to see how they cluster in normal populations, a discrete number of personality types are developed and often separated into symmetrically polar types (called a personality circumplex). Pathological personality types are seen as extremes of these conceptually organized normally occurring traits. In DSMIII, the avoidant category was derived from this conceptual/intellectual tradition.

Although personality variables may lend themselves to numerical representation, this is not sufficient justification for deciding on such a model for diagnostic purposes. For example, blood pressure is a continuous measure that is best displayed in numerical representation. However, a cutoff is chosen for diagnostic purposes to define "hypertension " so that this latter term communicates essential information about prognosis and treatment.

B. The dimensional model better demonstrates the relationship of personality disorders to traits occurring^ in the normal population. Frances (1982) has argued That dimensional models are superior to categorical models for those cases scoring at or close to the boundary between two categories, that is, normality versus abnormality. The assignment of such cases to categories greatly contributes to the problems of reliability and validity. If many patients are located at the perimeters of the categories, this argument is accurate. If most patients are prototypic for the categories, then it is not. In either event it can only be tested if and when prototypic cases are well defined.

C. The dimensional model has measurement advantages and increases reliability. A statistician would prefer a continuous or ratio variable over one that transforms these into nominal or categorical variables. Norman and Streiner (1986) insist that "classifying good ratio measures into large categories is akin to throwing away data." A dimensional model may improve the poor reliabilities (Mellsop, Varghese, Joshua, &i Hicks, 1982; Spitzer, Forman, Si Nee, 1979) that are observed when clinicians make Axis II diagnoses using the present DSM-III categories. However, the decision about a model should not be based only on psychometric properties but comes back to the purposes for the model. Elevated blood pressures may be best measured for statistical purposes as a continuous variable, but in terms of an outcome we may be most interested in the threshold at which lethal/nonlethal categories exist.

D. The dimensional model explains the problem of diagnostic overlap. This model has been used to explain the overlapping diagnoses observed whenever DSM-III criteria have been employed. Typically, psychiatric patients who fulfill criteria for any DSM-III personality disorder meet criteria for several others (Mellsop et al., 1982; Stangl et al., 1985; Zanarini et al., 1987 ). Advocates for the dimensional model believe that the high degree of overlap reflects accurately the true complexity of the subject. They argue that critics fail to appreciate that such overlap exists because "Maladaptive personality traits are extreme variants of normal traits that are not exclusive" (Widiger & Frances, 1985, p. 616 ). Overlapping personality diagnoses may simply mirror the complexity intrinsic to human personality; they are needed just as multiple personality traits are needed to describe normal individuals. However, patients with overlapping or multiple diagnoses may be the most severely disturbed individuals and/or may be very different from cases with specific diagnoses. For example, children with both attention deficit disorder and conduct disorder appear to be qualitatively different from children with one of these disorders alone (Szatmari, 1987). The issue of overlapping diagnoses is a research priority for all of psychiatry.

It is instructive to take the historical view and realize that the current overlaps in definitions of Axis II disorders are to some extent a repeat of the same developmental processes that occurred in the formulation of discrete Axis I categorizations. In the early 1970s, due to frequent concurrence, there was uncertainty about whether depression and anxiety were separate disorders, one disorder, or merely symptoms that should be measured as dimensional traits. Syndromes were identified in two fashions. First astute clinicians, through clinical experience and reviews of the literature, made clinically plausible formulations. An example of this would be Marks and Lader 's (1973) review of anxiety. Secondly, researchers utilized mathematical techniques to attempt to identify categories. Examples of this in the affective disorders are Derogates et al. (1972 ), Gurney et al. (1972 ), and Prusoff and Klerman (1974 ). These categorizations then were validated by methods such as course of illness, family studies, and eventually biochemical studies. Just as these Axis I categorizations were criticized for their overlap of symptoms, now the same nosologic developmental process is unfolding for the criteria for the personality disorders’ categorizations. First is description by skilled clinicians. Second is the definition of disorders by assembling large samples and differentiating them by statistical means from each other and from traits. Significant examples include Tyrer and Ferguson (1987) and Cloninger (1987). Similarly, attempts are being made at personality disorder validation by course of illness, family studies, and biological markers (eye tracking, REM latency, and responses to acoustical stimuli).

The second problem facing DSM-III personality disorders that has already occurred for Axis 1 disorders is co-occurrence. Although this has led to criticism that such personality disorders are not discrete entities, community epidemiologic surveys that indicate overlap of Axis I disorders (Myers et al., 1984; Robins et al., 1984) have not led to conclusions that the individual diagnoses are invalid. For Axis I disorders it simply raises interesting questions about the relationships between these disorders. Given that clinical populations have an even greater overlap, this overlap, in it self, is not a devastating criticism of categorical integrity. So it may be that what we are seeing in Axis II research is a repeat of developments seen for Axis I.

THE CATEGORICAL MODEL

Figure 1 also illustrates a categorical model. Like the dimensional model, the types it includes may have quantitative differences (e.g., levels of severity, prevalence rates, etc.) and have nearer or further neighbors. Unlike the dimensional model, the types are believed to have distinct boundaries de fined by qualitative differences. That is, they have defining, core, or essential characteristics that either qualify or disqualify an individual for the diagnosis.

Proponents offer the following four advantages for use of the categorical model:

A. Categorical models are familiar to clinicians and aid acceptance of a typofogy. Moreover...

Download PDF to continue reading article, footnotes, tables, and references.

James H. Reich, MD, MPH, is a board-certified Forensic Psychiatrist with extensive civil psychiatry experience who has done hundreds of evaluations. His services are available for civil and some criminal law cases. Clients are assured of his personal dedication to each case. He does high quality research, expert evaluations, writes a solid report, and will testify well. Dr. Reich has been deposed over 50 times.

©Copyright - All Rights Reserved

DO NOT REPRODUCE WITHOUT WRITTEN PERMISSION BY AUTHOR.