The dimensional view of personality disorders: a review of the taxometric evidence

Clinical Psychology Review 23 (2003) 75 – 93

The dimensional view of personality disorders: a review of the taxometric evidence Nick Haslam Department of Psychology, University of Melbourne, Parkville, VIC 3010, Australia Received 15 March 2002; accepted 7 August 2002

Abstract The dimensional view of personality disorders (PDs) represents these conditions as extreme variants of normal personality continua. This widely held view underpins efforts to characterize PDs in terms of established systems of personality description and to overhaul classification of PDs along dimensional lines. A review of 21 taxometric studies of PDs and related variables calls an unqualified version of this view into question. Analyses of the three PDs investigated to date strongly support taxonic (i.e., categorical or discontinuous) models. Implications for the conceptualization and classification of PDs are drawn. D 2003 Published by Elsevier Science Ltd. Keywords: Personality disorder; Dimensions; Categories; Taxometrics

1. Introduction Personality disorders (PDs) occupy an unusual position in psychiatric classification. Since introduction of DSM-III (American Psychiatric Association, 1980), they have been segregated from the great majority of mental disorders on a separate diagnostic axis. Their status as mental disorders has been questioned on many grounds, from pragmatic concerns about their diagnostic unreliability to more fundamental doubts about whether they represent diagnosable forms of abnormality.

E-mail address: [email protected] (N. Haslam). 0272-7358/03/$ – see front matter D 2003 Published by Elsevier Science Ltd. PII: S 0 2 7 2 - 7 3 5 8 ( 0 2 ) 0 0 2 0 8 - 8

76

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

One reason for the marginal and controversial status of PDs is that they have two aspects— personality and disorder—about which the clinical community tends to hold contrasting views. Mental disorders, following the standard medical model, are often understood to be discrete pathological entities. Individuals are generally taken either to have or not to have a particular disorder. Degrees of severity may be acknowledged among the affected and subthreshold forms may be recognized, but having the disorder is nevertheless understood in a categorical fashion. Consistent with this view, DSM-IV (American Psychiatric Association, 1994) makes categorical diagnoses, often defining a cutoff point on a symptom checklist above which a disorder is judged to be present. Personality is often conceptualized in an entirely different manner. Traditionally, variations in personality are understood to be matters of degree rather than kind. Differential psychologists pioneered a view of traits as quantifiable dimensions and dismissed typological accounts of personality such as Jung’s as coarse and prescientific. To this day, personality psychologists represent traits as graded continua and model their structure with statistical tools such as factor analysis that presume the adequacy of dimensional representations. The intellectual and historical context of this dimensional view of personality and of the categorical alternative are extensively discussed by Gangestad and Snyder (1985), Kagan (1994), and Meehl (1992). In the case of PDs these views collide. To the extent that a PD is a standard mental disorder it will be understood categorically, consistent with the default assumption embodied in DSM-IV. To the extent that a PD is a form of personality, however, a dimensional view seems appropriate, consistent with the representation of normal personality variation as continuous. Some of the controversies surrounding PDs are associated with these conflicting views. If PDs are grounded in normal personality continua, it seems questionable and arbitrary to conceptualize them as discrete categories, and the unreliability of their diagnosis may at least partly reflect this arbitrariness. The categorical versus dimensional status of PDs is not only an underlying basis for other controversies in the field, but has been a source of controversy in its own right since publication of DSM-III (Frances, 1982; Widiger & Clark, 2000). Many psychologists have argued for a dimensional view of PDs on a number of compelling grounds (e.g., Livesley, Schroeder, Jackson, & Jang, 1994; Widiger & Costa, 1994). First, a dimensional view may be truer to the fundamental nature of PDs than a categorical view, so that DSM-IV’s cutoffs impose arbitrary distinctions that misrepresent a seamless state of affairs. Second, by drawing arbitrary distinctions, the categorical view may reduce researchers’ capacity to assess the correlates of PDs by weakening statistical power, and may lose severity-related information that is important for clinicians. Third, the categorical view is held responsible for the high degree of ‘‘comorbidity’’ among PDs, so that individuals who fall high on a dimension common to several PDs are spuriously represented as having several distinct conditions. In short, proponents of the dimensional view criticize the categorical alternative as crudely black-and-white in its structural assumptions, oversimplifying and falsely precise in its dichotomization, and uneconomical in its diagnostic application. Instead of being discrete categories, they argue, PDs represent extreme variants that fall on a continuum with normal personality.

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

77

Advocates of the dimensional view can point to two main implications of their stance. First, the classification of PDs should be overhauled along dimensional lines, so that degrees of personality pathology can be assessed along a small number of nonredundant personality dimensions. Second, given the proposed absence of discontinuity between normal and abnormal personality, researchers can seek to understand PDs in terms of established models of normal personality. These two implications are obviously connected. If PDs were shown to be associated with particular trait dimensions, these dimensions might serve as bases for a radically revised dimensional classification of axis II. Attempts to understand PDs in terms of established dimensional models of normal personality are well advanced. The two most successful models used to date are the fivefactor model (FFM; John & Srivastava, 1999) and the interpersonal circle (IC; Wiggins, 1982). Numerous studies have mapped PDs onto these models, with substantial success (e.g., Lynam & Widiger, 2001; Romney & Bynner, 1997; Sim & Romney, 1990; Soldz, Budman, Demby, & Merry, 1993; Trull, 1992). These literatures have been reviewed elsewhere (Costa & Widiger, 1994; Widiger & Hagemoser, 1997), and their rapid growth is evident in a recent special issue of the Journal of Personality (Ball, 2001) and a second edition of Costa and Widiger’s (2001) volume on PDs and the FFM. Many PDs appear to be robustly correlated with the dimensions proposed by the respective dimensional models. At present, FFM, IC, and other dimensional systems of personality (e.g., Benjamin, 1996; Birtchnell & Shine, 2000; Cloninger, 1987) are widely believed to offer not only a valuable description of at least some PDs, but also an understanding of what constitutes or underlies them. On the dimensional view, PDs are elevations on particular trait continua, rather than simply being empirically associated with them. The dimensional view of PDs has garnered considerable support among psychologists. The evidence supporting it has been described as ‘‘overwhelming’’ (Livesley, 1996, p. 224) and most interested psychologists probably adhere to it explicitly or implicitly. However, just as the categorical view of PDs embodied in DSM-IV lacks empirical support and largely represents a taxonomic assumption, the evidential support for the dimensional view is not as uniform or substantial as some have claimed. For instance, evidence that PD symptom distributions are phenotypically continuous, lacking clear breaks or bimodalities (Livesley, Jackson, & Schroeder, 1992), is not inconsistent with the existence of latent categories whose score distributions overlap. Evidence that PD symptom scores are associated continuously with functional impairment (Nakao et al., 1992) also does not rule out presence of latent categories, as category members (and nonmembers) may have varying degrees of symptom severity. The extensive evidence that normal personality dimensions have reliably patterned associations with PDs does not establish that these PDs are merely extreme variants on these dimensions. What is needed to adjudicate between categorical and dimensional views of PDs more definitively is a rigorous method for comparing their adequacy as structural models of particular PDs. The taxometric procedures developed by Paul Meehl and his colleagues (e.g., Meehl, 1995; Waller & Meehl, 1998) are well suited to this task. Employed in roughly 70 empirical studies to date (see Haslam & Kim, in press, for a review), these procedures test between categorical (‘‘taxonic’’) and dimensional (‘‘nontaxonic’’) models by examining

78

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

patterns of covariation among fallible indicators of the latent variable of interest. Maximum covariance analysis (MAXCOV), for example, examines the covariance of a pair of indicators as a function of a third. Such covariance should rise to a maximum where the latent distribution of category members intersects the distribution of nonmembers (i.e., where members and nonmembers are in equal proportions), but should be flat where no latent category exists. Means-above-minus-below-a-cut (MAMBAC), the second most popular taxometric procedure after MAXCOV, has a similar rationale. The mean value of one indicator is calculated for cases falling above and below a sliding cutpoint on a second indicator, and the difference between these means should also rise to a peak where members and nonmembers of a latent category have equal mixing proportions. Taxometric methods such as these rely heavily on plot inspection rather than null hypothesis significance testing, and on the convergent findings of multiple independent procedures (‘‘consistency testing’’). Extensive Monte Carlo evidence shows that they discriminate taxonic and nontaxonic data sets with high accuracy (e.g., Meehl & Yonce, 1994, 1996), do so as well as or better than alternatives such as cluster analysis and mixture modeling (Cleland, Rothschild, & Haslam, 2000), and are reasonably robust under unfavorable measurement conditions such as skewed or dichotomous indicators (e.g., Haslam & Cleland, 1996; Ruscio, 2000). Taxometric studies have become popular in abnormal psychology, and have recently examined such clinical phenomena as depression (Ruscio & Ruscio, 2000), dissociation (Waller, Putnam, & Carlson, 1996), eating disorders (Gleaves, Lowe, Snow, Green, & Murphy-Eberenz, 2000), and worry (Ruscio, Borkovec, & Ruscio, 2001). Livesley et al. (1994) noted that the taxometric methods employed in studies such as these ‘‘have the potential to advance theoretical understanding of the nature of . . . personality disorder’’ (p. 10), and cited three PD-related papers, but since then many more studies have appeared. The present paper reviews these studies to determine where the rapidly accumulating evidence stands. Does it tend to support the dimensional view of PDs, the categorical view, or neither? A comprehensive search for published and unpublished research, conducted using extensive Internet, PsycLIT and bibliographic searches, and personal inquiries to investigators, revealed 21 taxometric studies of PDs or related variables. Studies of the three PDs that have received taxometric scrutiny are reviewed below, followed by a discussion of their implications.

2. Schizotypal PD (SPD) Taxometric interest in SPD arose in response to Meehl’s theory of schizophrenia (Meehl, 1962, 1990), which holds that genetic liability to the disorder is confined to a discrete class of people—‘‘schizotypes’’—who manifest a distinct form of personality organization. Meehl (1973) presented the rationale for MAXCOV to use fallible indicators of schizotypy to test the taxonic hypothesis. Using several such procedures, Golden and Meehl (1979) found strong convergent evidence for a taxon having a ‘‘pre-schizophrenic’’ profile in a sample of 211 nonpsychotic patients.

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

79

Since this first study of schizotypy, 11 more have been conducted. Two of these have examined samples at high risk of developing schizophrenia and nine have examined unselected student samples. In the first high-risk study, Erlenmeyer-Kimling, Golden, and Cornblatt (1989) found clear evidence of taxonicity in a nonclinical sample of 185 children, using standardized cognitive and neuromotor assessments. Forty-seven percent of children of a schizophrenic parent were identified as taxon members compared with 4% of normal controls, and taxon members were much more likely than nonmembers to have been admitted to a psychiatric hospital by young adulthood. Erlenmeyer-Kimling et al. (1989) replicated this finding of taxonicity in a second sample of 150. In the second high-risk taxometric study, Tyrka, Cannon, et al. (1995) and Tyrka, Haslam, and Cannon (1995) found a schizotypal taxon of stable base rate and composition in a longitudinal high-risk data set. At three waves of the study—from 311 participants’ mid-teens to late thirties—MAXCOV yielded consistently taxonic findings despite the use of distinctive assessments of schizotypy (most based on DSM-III-R SPD symptoms) at each wave, using interview and behavioral rating data. Taxon membership conferred a markedly increased risk of lifetime schizophrenia spectrum but not nonspectrum diagnosis, and was plausibly and strongly associated with risk status (i.e., having 0, 1, or 2 schizophrenic parents). Studies of college student samples have been conducted by several research groups. Aside from their dissimilar sample composition, they differ from the high-risk studies in having larger samples, using self-report psychometric scales, and often focusing attention on specific components of schizotypy or SPD rather than on the whole syndrome. Raulin and his colleagues (Lowrie & Raulin, 1990) conducted the first of these studies, and followed it up in another (Raulin, Lowrie, & Brenner, 1994; see also Brenner et al., 1995; Lowrie & Raulin, 1997). Assessing 2684 undergraduates on measures of perceptual aberration (i.e., perceptual distortions and disturbances of body image), magical ideation, and cognitive slippage, they repeatedly obtained taxonic findings using MAXCOV and MAMBAC. Taxon base rate estimates fell within a narrow range (0.07–0.10). The only exception to this otherwise highly consistent pattern was a nontaxonic finding when the MAXSLOPE procedure was performed (Raulin et al., 1994). This procedure (Grove & Meehl, 1993) has rarely been used in taxometric research, has little support from simulation studies, and appears to be somewhat unreliable. Three taxometric studies of college students (total N = 3068) were conducted by Lenzenweger and colleagues. In an analysis of perceptual aberration, Lenzenweger and Korfine (1992) obtained a taxonic finding with a base rate estimate of about 0.10. Korfine and Lenzenweger (1995) obtained a lower figure (0.03) in a replication study of perceptual aberration, while supporting taxonicity and showing that taxon members have elevated levels of schizotypal features as well as generalized personality disturbance. Broadening the range of indicator variables to include magical and referential thinking, Lenzenweger (1999) again obtained a taxonic result, this time with an estimated base rate of 0.13. Four further college student-based studies have been completed. MacFarlane (1996) replicated the taxonic status of two SPD-related indicators (perceptual aberration and social anhedonia), with base rates of about 0.10 in a sample of 1442, but obtained ambiguous findings for magical ideation and nontaxonic findings for physical anhedonia. Blanchard,

80

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

Gangestad, Brown, and Horan (2000) found social anhedonia to be taxonic with a base rate of 0.05 in a sample of 1526. Keller, Jahn, and Klein (2001) supported this claim with taxonic findings (base rate = 0.13; N = 1103) for SPD indicators emphasizing negative features (e.g., anhedonia-related constricted affect, lack of close friends, social anxiety) but obtained nontaxonic results for positive features (e.g., odd speech, ideas of reference, perceptual abnormalities). Finally, Meyer and Keller (2001) obtained taxonic results with base rates of around 0.12 (N = 809) for perceptual aberration and physical anhedonia but not magical thinking. All told, the 12 existing studies of SPD and related variables show a great deal of support for a categorical or taxonic view. Every study corroborates this view for at least one schizotypy-related component, and every such component has yielded a taxonic finding at least once, with perceptual aberration making the most consistent showing. The three nontaxonic exceptions (physical anhedonia, MacFarlane, 1996; positive symptoms, Keller et al., in press; magical ideation, Meyer & Keller, 2001) occur alongside 18 distinct analyses favoring taxonicity, and all three involve specific postulated components of schizotypy rather than the full syndrome. This specificity raises the possibility that the indicators used in the nonconforming analyses may have had insufficient validity to detect an SPD-related taxon individually, given that inadequate indicator validity is a primary source of spurious nontaxonic inference in taxometrics (Cleland et al., 2000). The nontaxonic findings might therefore reflect inadequacies in particular indicators, although it is also possible that the findings point accurately to elements of or related to schizotypy that are not themselves taxonic. In any event, consistency of taxonic findings in the reviewed studies is impressive, as are the remarkably convergent base rate estimates arising from the college student studies. This convergence manifests itself in spite of the studies’ varying indicator selection, measurement instruments, and research groups, and agrees strikingly with the 0.10 population prevalence rate predicted by Meehl (1990). It would be strange indeed for such a robust finding to emerge in the absence of a real SPD-related taxon.

3. Antisocial PD (APD) Five taxometric studies have addressed the status of APD or related variables. The first of these, an unpublished dissertation by Dana (1990), suggested that APD was nontaxonic. This study is significantly weakened by its failure to use any consistency tests to yield convergent support for its conclusion, and by its reliance on dichotomous indicators, whose threat to valid taxometric inference has been shown by Miller (1996) and Ruscio (2000). In addition, it relied exclusively on self-report assessment of APD-related features, using as indicators single MMPI items whose capacity to discriminate APD with sufficiently power (i.e., validity) is extremely questionable. The four remaining taxometric studies have yielded solidly taxonic findings. All employed at least one consistency test besides the primary taxometric procedure (M = 2.5). With one exception (Ayers, Haslam, Bernstein, Tryon, & Handelsman, 1999), all used continuous rather than dichotomous indicators, did not rely exclusively on self-report, and had larger

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

81

samples than Dana’s (N = 609). In a comprehensive analysis using particularly careful assessments, Harris, Rice, and Quinsey (1994) found convergent evidence for taxonicity using indicators of psychopathy. Although psychopathy is not entirely redundant with APD, its overlap is substantial, and interestingly, Harris et al. (1994) determined that it was the APD-like aspects of the broader psychopathy construct—those representing chronic antisocial behavior beginning in childhood—that were most clearly taxonic. The interpersonal and affective aspects of psychopathy, as well as adult criminality per se, were nontaxonic. Ayers et al. (1999) supported this finding of taxonicity in a study of APD symptoms in 653 poly-substance abusing adults. Skilling, Harris, Rice, and Quinsey (2001) found evidence of the same taxon in a sample of 440 persistently antisocial adult offenders, using measures of psychopathic tendencies and DSM-IV APD indicators. Skilling, Quinsey, and Craig (2001) detected a taxon of antisocial boys of middle-school age in a community sample of 1111, implying that the class identified in previous studies does not originate in adult criminal careers, consistent with Harris et al.’s finding about the discreteness of chronic antisociality. Dana’s (1990) relatively weak study sounds a dissenting note, but the evidence of four stronger and more recent studies supports the taxonicity of a latent variable corresponding well to APD. Although Skilling et al. (2001) did not directly examine APD, using a child rather than an adult sample, their finding is consistent with existence of an APD taxon, given that the disorder is conceptualized as having childhood conduct problems as necessary precursors. In view of the oft-remarked failure of recent DSM editions of APD criteria to encompass psychopathy satisfactorily, it is interesting that evidence for this taxon is strongest precisely in those aspects of the psychopathy construct that DSM embodies, namely, antisocial behavior with a history of childhood conduct problems.

4. Borderline PD (BPD) Four taxometric studies of BPD have been conducted. These show less agreement in their findings than the studies of SPD and APD. In the first study, Trull, Widiger, and Guthrie (1990) obtained clinical ratings of BPD symptoms of 409 psychiatric outpatients by careful chart review. MAXCOV analysis yielded a monotonically increasing plot that the authors interpreted as evidence of continuity, on the mistaken view that taxonic plots are always centrally peaked. As others (e.g., Korfine & Lenzenweger, 1995; Meehl, 2001) have noted, this increasing plot is much more likely to reflect a taxon with a low base rate, as a borderline taxon would be expected to have in an outpatient sample (i.e., 0.10; American Psychiatric Association, 1994). However, two later taxometric studies using MAXCOV—Ayers et al. (1999) with 440 poly-substance abusing adults and Simpson (1994) with 552 outpatients— obtained nontaxonic findings for BPD that do not suffer from this misinterpretation. The findings of these three studies are inconsistent and ambiguous, and they are open to question on several counts. First, the Trull et al. (1990) study should probably be given the most weight, as its indicator variables were based on clinical judgments from a thorough chart review. The Simpson (1994) and Ayers et al. (1999) studies, in contrast, employed a selfreport measure of BPD symptoms that is notable for its transparent item content (PDQ-R/

82

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

PDQ-4; Hyler, 1994). It is questionable indeed that this instrument, whose scales correlate poorly with more reliable interview-based PD assessments (Widiger & Sanderson, 1995), allows a sufficiently valid assessment of BPD symptoms, especially as single-item indicators were used in both studies. Monte Carlo simulations (e.g., Cleland et al., 2000) show that low indicator validity weakens the power of taxometric procedures to detect taxa, thereby reducing confidence in Simpson’s and Ayers et al.’s nontaxonic findings. A second set of weaknesses reduces confidence in all three BPD studies. Their taxometric findings are undermined by modest sample sizes, insufficient consistency testing, and the use of dichotomous indicators. Although their samples all exceed the minimum of 300 recommended by Meehl, larger samples may be needed to detect low base rate taxa such as BPD. Taxometric procedures have shown themselves to be relatively insensitive to taxa with base rates such as these (Cleland et al., 2000), needing larger samples to compensate. Such lack of power to detect low base rate taxa is compounded by the use of dichotomous indicators, which Ruscio (2000) showed to decrease the reliability of taxometric inference. Ruscio proposed a minimum N of 600 when dichotomous indicators are used, which all three BPD studies fail to meet. Finally, consistency testing is a cardinal feature of taxometric analysis, but between them the three studies used only two such tests. Ayers et al. (1999) employed MAMBAC, Simpson (1994) used a data simulation procedure, and Trull et al. (1990) employed no consistency testing. In sum, empirical foundations of the existing taxometric studies of BPD are somewhat shaky, with the scales of evidence perhaps tipped slightly towards taxonicity by the superior data quality of Trull et al. (1990). A more recent study by Rothschild, Cleland, Haslam, and Zimmerman (submitted for publication), however, avoided many of the pitfalls of its predecessors. This study used a much larger sample (N = 1395), structured interview-based assessment of BPD symptoms, three consistency tests, and indicators that were continuous and multi-item-based rather than dichotomous and single-item. The study yielded strong support for the taxonicity of BPD, three distinct analyses (MAXCOV, MAMBAC, and mixture modeling) yielding plainly taxonic results and highly convergent base rate estimates. Although replication is desirable, Rothschild et al.’s (submitted for publication) study places the categorical view of BPD on a firmer empirical footing than the dimensional view, particularly when the next most compelling study (Trull et al., 1990) yields the same conclusion.

5. Summing up I have briefly reviewed the findings of 21 studies employing 22,689 participants. These studies have yielded a total of 30 distinct taxometric tests of particular PDs or PD-related variables—not counting consistency tests performed on the same data as the primary test— given that some studies examined more than one variable, used a replication sample (Erlenmeyer-Kimling et al., 1989), or analyzed longitudinal data at multiple waves (Tyrka et al., 1995). Twenty-four (80%) of these 30 tests have yielded taxonic findings. The six nontaxonic exceptions have tended to be analyses of specific components of a PD rather than

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

83

the complete syndrome (i.e., SPD), or have substantial methodological weaknesses. The 15 studies yielding exclusively taxonic findings do indeed tend to be methodologically stronger than the six studies yielding mixed or nontaxonic findings (i.e., Ayers et al., 1999; Dana, 1990; Keller et al., 2001; MacFarlane, 1996; Meyer & Keller, 2001; Simpson, 1994). The latter, four of which report unpublished dissertation studies, tend to have smaller N (M = 824.5 vs. 1212.1), are less likely to employ continuous indicators (33.3% vs. 60%), and use fewer consistency tests in addition to the primary taxometric procedure (M = 0.83 vs. 2.13). Only the last difference reaches statistical significance [t(19) = 2.28, P < .05], but the overall trends are clear. In sum, the taxometric evidence is highly consistent in its thrust, favoring a taxonic understanding of the three PDs thus far investigated, with support strongest for SPD and weakest for BPD. These findings cast considerable doubt on the dimensional view’s adequacy as a complete or general account of PDs. Although only 3 of the 10 recognized PDs have received taxometric attention, current evidence clearly does not support a presumption of dimensionality for the seven remaining PDs. Taxometric research on these disorders is urgently required. It may eventually support dimensional models for all of these disorders, but the research conducted to date suggests that this is an unlikely possibility. To expect consistent support for the dimensional view where these seven disorders are concerned, after the examined three have contradicted it, is to cling to a theoretical preference in the face of the evidence. Taxometric research strongly suggests that a categorical view is plausible for at least some PDs. This review does not advocate a wholesale categorical view of PDs, but simply suggests that the appropriate structural model for particular PDs should be an open empirical question, not a settled assumption.

6. Possible objections Several objections to this conclusion should be anticipated. For a start, it can be asked whether taxometric methods have an inbuilt bias towards taxonic findings, and therefore prejudge the categorical versus dimensional status of PDs. This concern can be allayed by two arguments. First, Monte Carlo studies of taxometric procedures consistently show no bias towards either taxonic or nontaxonic findings. Nontaxonic MAMBAC and MAXCOV plot panels are no more likely to be judged taxonic than are taxonic panels to be judged nontaxonic. The error rate of both procedures is symmetrical in this sense (Cleland et al., 2000). This fact can be contrasted to the inbuilt biases towards taxonic findings for cluster analysis and towards nontaxonic findings for factor analysis, and the weaker but still real bias towards nontaxonic findings in mixture models (e.g., Everitt & Hand, 1981; MacLean, Morton, Elston, & Yee, 1976), where nontaxonic models are retained by default unless taxonic models can demonstrate statistically superior fit. Taxometric methods as commonly employed may even have a weak bias in the same direction, because when multiple procedures are used (i.e., consistency testing), any disagreement among them may lead researchers to settle for the ‘‘conservative’’ or ‘‘parsimonious’’ conclusion of nontaxonicity. Second, taxometric methods have now been applied to 60 latent variables in psychology (see

84

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

Haslam & Kim, in press, for a review) and exactly half have yielded taxonic conclusions. Taxometric methods therefore do not inappropriately favor taxonic findings. Another possible objection to the conclusions of this review is that the three PDs studied thus far are in some respects unrepresentative of axis II. If this were the case, any generalized conclusion about the categorical versus dimensional status of PDs from these unusual examples would be risky. In response, it should be emphasized that this review is not arguing that PDs in general are categorical, but only that they cannot be confidently assumed to be completely dimensional on the available evidence. However, it is conceivable that SPD, APD, and BPD are atypical in some sense. Arguably, they are more disabling and severe than most PDs, and to the extent that severity is associated with taxonicity—an association that laypeople appear to believe (Haslam, 2000; Haslam & Ernst, in press)—the remaining PDs may be more apt to be nontaxonic. Alternatively, the three PDs may be those that clinicians judge to be the best candidates for inclusion in axis I.1 Nevertheless, the fact remains that all three disorders studied to date are currently considered to be PDs, all have been modeled by proponents of the dimensional view in terms of personality continua such as the FFM, and all appear to be taxonic. What makes SPD, APD, and BPD atypical may simply be that they have attracted more research attention than the others. There seems little a priori reason to expect that the categorical versus dimensional status of the seven more neglected PDs should be distinctly different from the more extensively studied minority. A more potent objection is that the taxometric studies reviewed above have not always directly addressed the PDs in question, but have sometimes investigated latent variables that are merely related. This objection is difficult to sustain for the studies of BPD, which have all employed as indicators either DSM diagnostic criteria or self-report scales explicitly assessing the disorder. The objection is more defensible in relation to APD, where some taxometric studies have used indicators of psychopathy or childhood antisociality rather than adult APD. However, in this case, the critique is mitigated by the fact that it was the specifically APDrelated components of psychopathy that were most clearly taxonic, and by the conceptualization of APD as having developmental continuity with antisociality prior to adulthood. It should also be remembered that the studies that used direct measures of APD also yielded taxonic findings, implying that the conclusion of taxonicity cannot be explained away by the employment of indicator variables that are not APD-specific in some studies. The case of SPD is less straightforward. Some taxometric studies directly assessed SPD symptoms as enumerated in recent versions of DSM (e.g., Tyrka, Cannon, et al., 1995; Tyrka, Haslam, et al., 1995), and obtained taxonic findings. However, most studies did not, using instead measures of constructs such as perceptual aberration, magical ideation, and physical and social anhedonia. It could be objected that these constructs do not adequately encompass SPD. This objection is mostly spurious. Indicators in a taxometric study do not need to ‘‘encompass’’ the construct whose categorical versus dimensional status they are assessing, but need only to be associated with it strongly. For example, the distribution of height or voice pitch in the adult population could be used to show that biological sex is taxonic, both being valid indicators of this variable, without any implication that they ‘‘encompass’’ 1

I thank an anonymous reviewer for this observation.

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

85

biological sex. Similarly, measures of perceptual aberration or social anhedonia do not encompass the criteria for SPD, but they bear a strong relation to existing diagnostic criteria, and have long been considered valid components of schizotypy. Thus, studies focused on such single measures can still be considered tests of the taxonicity of schizotypy more broadly, to the extent that the measures are indeed validly indicators. In any case, when multiple such measures have been used in taxometric analyses, allowing a more comprehensive coverage of schizotypy (i.e., the work of Raulin et al.), taxonic findings have also been obtained. A related concern is that specific aspects of schizotypy such as perceptual aberration and magical ideation do not always intercorrelate strongly, and that findings of taxonicity for single aspects may therefore not be informative about the full syndrome. This concern can be answered in several ways. First, although they are not always strongly associated, these aspects nevertheless cohere in classical and DSM representations of schizotypy and SPD, indicating that they do form a broader syndrome. Second, modest correlations among scales for the different aspects of schizotypy are expectable under the taxonic model. Given that associations among these scales are based primarily on the existence of a low-prevalence high-scoring taxon under this model, these correlations would be attenuated by the very large majority of nontaxon cases, among whom the measures would not be expected to have any strong association. Third, the great consistency of base rate estimates that taxometric studies of nonclinical samples have obtained, regardless of which aspect of schizotypy they have investigated, is most parsimoniously accounted for by positing a single schizotypal taxon that all aspects identify. A more pointed concern accepts that the taxometric studies reviewed above demonstrate that schizotypy is taxonic, but not that SPD is taxonic. This concern clearly has some merit. If the base rate of the taxon in reasonably representative population samples hovers around 0.10, and the population prevalence of SPD is estimated at a considerably lower rate, then somewhat different sets of people are being picked out by the taxon and the diagnosis. The divergent base rates of the taxon and diagnosed SPD might be explained in two ways. First, it could be argued that SPD’s diagnostic criteria are too conservative and the disorder is underdiagnosed. Second, it could be argued that the SPD diagnosis is appropriately restricted to a subset of taxon members whose schizotypal features are particularly severe or needful of clinical attention. What surely cannot be disputed is that people diagnosed with SPD would have a very strong tendency to be taxon members, and probably represent a nested subset of the taxon, given that the taxon is defined by the same features that distinguish SPD. The two explanations sketched above differ in their implications for the taxonicity of SPD. According to the first, SPD is taxonic because it ought to be identified with the broader schizotypal taxon. According to the second, SPD may be nontaxonic because it represents a relatively severe variant within the taxon. However, this second explanation offers little support for the dimensional view of SPD for several reasons. First, even if SPD is a relatively severe variant within the schizotypal taxon, it might well be a taxonic variant or ‘‘subtype.’’ Studies conducted to date have not set out to test this possibility nor have they been capable of doing so. Second, even if SPD were shown to fall on a severity continuum within the taxon, this would hardly vindicate the dimensional view of SPD. SPD would still be substantially grounded in a pathological category—one associated with SPD symptoms

86

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

and increased risk for schizophrenia-spectrum disturbance—and this category would need to be reckoned within any attempt to make sense of the disorder. To treat SPD as dimensional under these circumstances would be to ignore a fundamental discontinuity that underpins it. Proponents of the dimensional view of SPD are surely proposing that the disorder falls on a continuum with normality, not that it falls on a continuum within an abnormal category.

7. The PD findings in the context of other taxometric research If taxometric research calls an unqualified dimensional view of PDs into question, does it also make the categorical view more intelligible? A review of taxometric studies suggests that it does. Twenty-four studies of nonpathological personality variables have been conducted to date (see Haslam & Kim, in press), and most favor the dimensional view. Taxonic findings were obtained for only seven variables (29% of the total), these exceptions being sexual orientation (Gangestad, Bailey, & Martin, 2000), self-monitoring (Gangestad & Snyder, 1985), infant reactivity (Woodward, Lenzenweger, Kagan, Snidman, & Arcus, 2000), hypnotic susceptibility (Oakman & Woody, 1996), type A (Strube, 1989), and certain response styles (Strong, Greene, & Schinka, 2000). Three studies of broad personality dimensions such as the FFM traits (Arnau, Green, & Tubre, 1999; Green, Arnau, & Gleaves, 1999) and Jung’s preferences (e.g., feeling vs. thinking; Arnau, Green, Rosen, Gleaves, & Melancon, in press) have consistently supported their dimensionality, and three studies of personality characteristics conceptualized as diatheses (hypomanic temperament, Meyer & Keller, in press; anxiety susceptibility, Taylor, Rabian, & Fedoroff, 1999; and vulnerability to alcoholism, Knowles, 1989) have also yielded nontaxonic findings. In short, it is not at all unreasonable to assume that PDs might be dimensional, given that normal personality, broad traits used to model PDs, and even personality traits associated with mental disorders, are also typically dimensional. It is simply an assumption that the existing taxometric evidence fails to support. Numerous taxometric studies of mental disorders other than PDs have also been conducted (see Haslam & Kim, in press). Of these 20 conditions—axis I disorders or their proposed subtypes—a narrow majority (55%) have yielded taxonic findings, almost twice the rate for normal personality variables. The common assumption that axis I conditions are well understood categorically receives qualified support in this finding. The taxometric evidence indicates that taxonic variables are more commonly found in the domain of psychopathology than in the domain of normal personality, but also suggests that at least some PDs follow the former domain’s pattern more than the latter’s. It may be that structurally, PDs have more in common with psychopathology than with normal personality. In this respect, they may be less peculiar or unrepresentative forms of psychopathology than some have thought.

8. Implications for the dimensional view of PDs Earlier I summarized the flaws that proponents of the dimensional view of PDs saw in the categorical view. The categorical view was structurally inaccurate, drawing crude binary

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

87

distinctions. By dichotomizing cases and noncases, it made arbitrary judgments that jettisoned clinically and statistically important variance. By failing to recognize shared dimensional components of personality disturbance, it yielded spurious ‘‘comorbidity.’’ How do these criticisms fare in the light of this review? On the first point, the review indicates that the categorical view is not uniformly inaccurate as a structural account of PDs. The three PDs discussed were better modeled by two latent distributions—corresponding to taxon members and nonmembers—than by a graded continuum. However, it is a misunderstanding to think that such taxonic models are necessarily crude. Taxonic models do not imply that at the manifest level members can be crisply distinguished from nonmembers, just that at the latent level categorically distinct classes exist. Moreover, as Waller and Meehl (1998, p.9) note, there will generally be dimensional variation within the taxon class and its complement, so the ‘‘taxonic-versus-dimensional’’ distinction might more strictly be referred to as ‘‘taxonic dimensional versus dimensional only.’’ Variation within taxon and complement classes is not denied or disregarded in taxonic models. Finally, taxonic models need not even be strictly binary. A latent variable may be taxonic in the sense of having discrete taxon and complement classes but also have an intermediate class that resembles the taxon class by degrees. Such ‘‘fuzzy categories’’ have been the focus of some recent simulation research (Haslam & Cleland, 2002). Rothschild et al. (submitted for publication) found that a fuzzy model, incorporating taxon and complement classes and an intermediate class containing dimensional variation, fit their BPD data at least as well as a standard two-class taxonic model. In short, the categorical view of PDs need not be coarse or black-and-white. The second criticism of the categorical view, that it draws arbitrary distinctions and loses important information about levels of PD symptoms, can also be partially deflected. It should be noted that the first part of this criticism presumes that PDs are truly dimensional. If at least some PDs are better understood as categories, then there is in principle a category boundary that a diagnostic cutoff could approximate in an entirely nonarbitrary way. A category has a determinate base rate which a dimensional extreme does not. Diagnostic cutoffs could therefore be calibrated to maximize correct identification of taxon members and nonmembers. Rothschild et al. (submitted for publication), for instance, found that identification of borderline taxon members was optimized by relaxing the DSM-IV cutoff from five to four criteria out of nine, suggesting that the DSM-IV cutoff is too conservative. The loss of information criticism is more reasonable, and a purely dichotomous diagnosis of PDs may indeed lose information that weakens statistical power and clinical prediction. However, as the previous paragraph noted, the categorical view does not imply the absence of meaningful variation within the taxon or complement classes. Taxon members will typically differ in the severity of their condition in ways that may be predictively valid, and nonmembers may also differ informatively in their resemblance to the taxon. The categorical view of PDs is therefore quite consistent with recognizing the importance of continuous variation in PD symptomatology, so long as this variation is not taken to be the only kind involved. The empirical fact that continuous PD scales may outperform dichotomous PD diagnoses in predicting clinical phenomena (e.g., Nakao et al., 1992; Ullrich, Borkenau, & Marneros, 2001) is therefore not good evidence against the categorical view of PDs.

88

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

The third main criticism of the categorical view of PDs has to do with comorbidity. This review has little to say directly about this issue, except with respect to the three PDs discussed. It suggests that any comorbidity of these PDs is likely to indicate real co-presence of distinct disorders rather than spurious resemblance based on shared personality characteristics. If the three PDs are discrete categories and are not identical, then an individual correctly diagnosed with two or more of them truly has coexisting PDs (Meehl, 2001). This may not be the case if other PDs prove to be nontaxonic. However, it does suggest that PD comorbidity cannot be attributed exclusively to dimensional overlap or treated as convincing evidence that something is wrong with the diagnostic classification of PDs. The high levels of PD comorbidity that are often obtained may eventually prove to be exaggerated by redundancies among nontaxonic PDs, but some true comorbidity of discrete conditions is likely to remain. This comorbidity (e.g., of APD and BPD) may be based on common factors, just as pneumonia and frostbite may result from prolonged exposure to intense cold, but this causal commonality does not invalidate the discreteness of the two disorders.

9. Research implications If it were shown that most PDs were indeed in some respects categorical, in line with the taxometric evidence regarding APD, BPD, and SPD, should researchers abandon attempts to model PDs in terms of dimensional personality systems such as the FFM and the IC? The answer to this question is clearly ‘‘no.’’ First, these systems plainly describe fundamental and broadly influential dimensions of psychological variation, and it would be surprising indeed for any personality taxon to be unrelated to them. The FFM, IC, and perhaps, other systems should therefore afford at least partial characterizations of any PDs that prove to be latent categories, and should therefore sharpen the description of these PDs, especially when more fine-grained traits such as FFM facets are also employed (e.g., Reynolds & Clark, 2001). Second, associations between standard personality dimensions and taxonic PDs might go beyond description and point to causal relations. For instance, a PD might be linked to a personality dimension by some form of threshold effect, affecting people who fall beyond some value on the dimension. Alternatively, the elevation of a taxonic PD on a dimension might indicate a process that is specifically disturbed among taxon members (e.g., emotional dysregulation reflected by high neuroticism in BPD). The well-established fact that PDs correlate strongly and consistently with normal trait dimensions therefore does not entail that PDs are intrinsically dimensional or that they are ‘‘nothing but’’ extreme variants on these dimensions. In short, even if the categorical view of PDs was to receive more widespread support than it has to date, there would be no reason to discontinue research on their associations with normal personality continua. Usefulness of such continua for describing PDs is now beyond doubt. It would only be necessary to remember that any such associations would leave unanswered the important question of what produced the taxon’s discontinuity with normal personality, given that normal personality appears to be generally continuous, and that they might point to descriptive rather than causally relevant features of the PDs. Investigators

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

89

would need to acknowledge that although these associations may elucidate PDs, they cannot tell the whole story of what a taxonic PD is or how it arose as a taxon.

10. Implications for the classification of PDs The apparent taxonicity of PDs examined to date should dampen some of the enthusiasm for dimensional systems of PD classification, at least for the PDs concerned. The taxometric evidence suggests that there are nonarbitrary category boundaries separating affected and unaffected individuals, and that diagnostic systems should aim to pick out these boundaries. Individuals on different sides of a well-calibrated diagnostic rule for these PDs should tend to differ in clinically important ways. However, the fact of taxonicity does not negate the desirability of assessing degrees of severity among taxon members or degrees of resemblance to the taxon among nonmembers. Even if BPD is a category, some cases will be more prototypically borderline than others, and noncases will vary consequentially in their affective instability. In short, it may be ideal to develop a diagnostic system for axis II that combines categorical and dimensional features rather than opting for one or the other. Categorical distinctions appear to be real within this domain, and to ignore them or rule them out a priori is as counterproductive as insisting that a dichotomous diagnostic judgment is clinically sufficient. Such a combined approach also has the advantage of comprehensiveness. Given that it seems likely that some PDs will be taxonic and others nontaxonic, a combined categorical/dimensional diagnostic system avoids the cumbersomeness of having a categorical diagnosis for some PDs and a dimensional diagnosis for others.

11. Conclusion Whether PDs should be classified dimensionally or categorically is a complex question with many facets. It is not simply a question of the structural validity of the alternative models, but also involves pragmatic considerations such as clinical convenience, parsimony, and predictive utility (Frances, 1982). As a result, taxometric evidence cannot resolve the question outright. Nevertheless, taxometric analyses offer particularly rigorous examinations of the structural basis of PDs that cannot be ignored by theorists and researchers interested in the issue of categorical versus dimensional diagnosis. Taxometric evidence favoring a categorical view of certain PDs needs to be taken into account in any forming consensus on this issue. Such evidence will need to be integrated with evidence favoring dimensionality, as neither alternative is likely to supersede the other completely. The taxometric research reviewed in this paper covers only three PDs and therefore does not allow any strong generalizations about the status of PDs as a set. The limited coverage of this research, and the heterogeneity of its measures, methods, and samples, sounds an urgent call for extensions and replications. However, the research that is now at hand disallows blanket claims that PDs are always and utterly dimensional. This position may prove to be accurate for some PDs, pending future work, but current evidence strongly suggests that it is

90

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

inaccurate for several and that no firm expectations to the contrary are therefore justified for the rest. The categorical view of PDs cannot be completely dismissed at this point, and its tenability should be judged in part on the sort of empirical grounds that taxometric research provides. This view does not rule out the reality or predictive value of continuous variation in PD symptomatology, nor does it entail a crudely dichotomous understanding of PD categories. It can therefore be reconciled with some aspects of the dimensional view, especially where the supplementation of dichotomous diagnosis with quantitative assessment of severity is concerned. Generalized and unqualified theoretical positions on the dimensional versus categorical issue are unhelpful, and some kind of hybrid or pluralistic understanding of PDs is desirable. References American Psychiatric Association (1980). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: Author. American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. Arnau, R. C., Green, B. A., Rosen, D. H., Gleaves, D. H., & Melancon, J. G. (in press). Do Jungian types and dichotomous preferences really exist? An empirical investigation using taxometric analysis. Personality and Individual Differences. Arnau, R. C., Green, B. A., & Tubre, T. (1999). Taxometric analysis of Goldberg’s five-factor markers. Poster presented at the annual meeting of the Eastern Psychological Association, Providence, RI. Ayers, W., Haslam, N., Bernstein, D., Tryon, W. T., & Handelsman, L. (1999). Categorical vs. dimensional models of personality disorder in substance abusers: a taxometric analysis. Poster presented at the Annual Meeting of the American Psychiatric Association. Ball, S. A. (2001). Reconceptualizing personality disorder categories using personality trait dimensions: introduction to special edition. Journal of Personality, 69, 147 – 153. Benjamin, L. S. (1996). Interpersonal diagnosis and treatment of personality disorders (2nd ed.). New York: Guilford. Birtchnell, J., & Shine, J. (2000). Personality disorders and the interpersonal octagon. British Journal of Medical Psychology, 73, 433 – 448. Blanchard, J. J., Gangestad, S. W., Brown, S. A., & Horan, W. P. (2000). Hedonic capacity and schizotypy revisited: a taxometric analysis of social anhedonia. Journal of Abnormal Psychology, 109, 87 – 95. Brenner, V., Raulin, M. L., & Young, B. L. (1995). Comparing the sensitivity of two taxometric techniques in detecting the schizotypal taxonomy. Poster presented at the Annual Meeting of the Society for Research in Psychopathology, Iowa City, IA. Cleland, C., Rothschild, L., & Haslam, N. (2000). Detecting latent taxa: Monte Carlo comparison of taxometric, admixture and clustering procedures. Psychological Reports, 87, 37 – 47. Cloninger, C. R. (1987). A systematic method for clinical description and classification of personality variants. Archives of General Psychiatry, 44, 573 – 588. Costa, P. T., & Widiger, T. A. (Eds.) (1994). Personality disorders and the five-factor model of personality. Washington, DC: American Psychological Association. Costa, P. T., & Widiger, T. A. (Eds.) (2001). Personality disorders and the five-factor model of personality (2nd ed.). Washington, DC: American Psychological Association. Dana, D. A. (1990). Alcoholism and antisocial personality disorder: Related discrete personality variables? Unpublished doctoral dissertation, University of Montana. Erlenmeyer-Kimling, L., Golden, R. R., & Cornblatt, B. A. (1989). A taxonometric analysis of cognitive and neuromotor variables in children at risk for schizophrenia. Journal of Abnormal Psychology, 98, 203 – 208.

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

91

Everitt, B. S., & Hand, D. J. (1981). Finite mixture distributions. New York: Chapman & Hall. Frances, A. (1982). Categorical and dimensional systems of personality diagnosis: a comparison. Comprehensive Psychiatry, 23, 516 – 527. Gangestad, S. W., Bailey, J. M., & Martin, N. G. (2000). Taxometric analyses of sexual orientation and gender identity. Journal of Personality and Social Psychology, 78, 1109 – 1121. Gangestad, S. W., & Snyder, M. (1985). ‘‘To carve nature at its joints’’: on the existence of discrete classes in personality. Psychological Review, 92, 317 – 349. Gleaves, D. H., Lowe, M. R., Snow, A. C., Green, B. A., & Murphy-Eberenz, K. P. (2000). Continuity and discontinuity models of bulimia nervosa: a taxometric investigation. Journal of Abnormal Psychology, 109, 56 – 68. Golden, R. R., & Meehl, P. E. (1979). Detection of the schizoid taxon with MMPI indicators. Journal of Abnormal Psychology, 88, 217 – 233. Green, B. A., Arnau, R. C., & Gleaves, D. H. (1999). Taxometric analysis of the five-factor model constructs. Poster presented at the annual meeting of the American Psychological Association, Boston, MA. Grove, W. M., & Meehl, P. E. (1993). Simple regression-based procedures for taxometric investigations. Psychological Reports, 73, 707 – 737. Harris, G. T., Rice, M. E., & Quinsey, V. L. (1994). Psychopathy as a taxon: evidence that psychopaths are a discrete class. Journal of Consulting and Clinical Psychology, 62, 387 – 397. Haslam, N. (2000). Psychiatric categories as natural kinds: essentialist thinking about mental disorders. Social Research, 67, 1031 – 1058. Haslam, N., & Cleland, C. (1996). Robustness of taxometric analysis with skewed indicators: II. A Monte Carlo study of the MAXCOV procedure. Psychological Reports, 79, 1035 – 1039. Haslam, N., & Cleland, C. (2002). Taxometric analysis of fuzzy categories: a Monte Carlo study. Psychological Reports, 90, 401 – 404. Haslam, N., & Ernst, D. (in press). Essentialist beliefs about mental disorders. Journal of Social and Clinical Psychology. Haslam, N., & Kim, H. (in press). Categories and continua: a review of taxometric research. Genetic, Social, and General Psychology Monographs. Hyler, S. E. (1994). PDQ-4 manual. New York: NiJo Publications. John, O. P., & Srivastava, S. (1999). The big five trait taxonomy: history, measurement, and theoretical perspectives. In L. A. Pervin (Ed.), Handbook of personality: theory and research, vol. 2 (pp. 102 – 138). New York: Guilford. Kagan, J. (1994). Galen’s prophecy. New York: Basic. Keller, F., Jahn, T., & Klein, C. (2001). Anwendung von taxometrischen methoden und von mischverteilungsmodellen zur erfassung von schizotypie (Application of taxometric methods and mixture distribution models for the assessment of schizotypy). In B. Andresen, & R. Mass (Eds.), Schizotypie. Psychometrische entwicklungen und biopsychologische forschungsansa¨tze ( pp. 391 – 412). Go¨ttingen: Hogrefe. Knowles, E. E. (1989). Detection of a latent alcohoid taxon: a taxometric search. Unpublished doctoral dissertation, University of Arkansas. Korfine, L., & Lenzenweger, M. F. (1995). The taxonicity of schizotypy: a replication. Journal of Abnormal Psychology, 104, 26 – 31. Lenzenweger, M. F. (1999). Deeper into the schizotypy taxon: on the robust nature of maximum covariance analysis. Journal of Abnormal Psychology, 108, 182 – 187. Lenzenweger, M. F., & Korfine, L. (1992). Confirming the latent structure and base rate of schizotypy: a taxometric analysis. Journal of Abnormal Psychology, 101, 567 – 571. Livesley, W. J. (1996). Commentary on ‘‘Epistemic value commitments’’. Philosophy, Psychiatry, and Psychology, 3, 223 – 226. Livesley, W. J., Jackson, D. N., & Schroeder, M. L. (1992). Factorial structure of traits delineating personality disorders in clinical and general population samples. Journal of Abnormal Psychology, 101, 432 – 440. Livesley, W. J., Schroeder, M. L., Jackson, D. N., & Jang, K. L. (1994). Categorical distinctions in the study of personality disorder: implications for classification. Journal of Abnormal Psychology, 103, 6 – 17.

92

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

Lowrie, G. S., & Raulin, M. L. (1990). Search for schizotypic and nonschizotypic taxonomies in a college population. Paper presented at the Sixty-First Annual Convention of the Eastern Psychological Association, Philadelphia, PA. Lowrie, G. S., & Raulin, M. L. (1997). Testing the construct validity of Meehl’s taxonic model of schizotypy: taxometric search techniques applied to self-report measures of schizotypic signs. Unpublished manuscript, State University of New York at Buffalo. Lynam, D. R., & Widiger, T. A. (2001). Using the five factor model dimensions to represent the DSM-IV personality disorders: an expert consensus approach. Journal of Abnormal Psychology, 110, 401 – 412. MacFarlane, R. (1996). Taxometric analysis of schizotypy. Unpublished doctoral dissertation, University of Iowa. MacLean, C. J., Morton, N. E., Elston, R. C., & Yee, S. (1976). Skewness in commingled distributions. Biometrics, 32, 695 – 699. Meehl, P. E. (1962). Schizotaxia, schizotypy, schizophrenia. American Psychologist, 17, 827 – 838. Meehl, P. E. (1973). MAXCOV-HITMAX: a taxonomic search method for loose genetic syndromes. In P. E. Meehl (Ed.), Psychodiagnosis: selected papers ( pp. 200 – 224). Minneapolis, MN: University of Minnesota Press. Meehl, P. E. (1990). Toward an integrated theory of schizotaxia, schizotypy, and schizophrenia. Journal of Personality Disorders, 4, 1 – 99. Meehl, P. E. (1992). Factors and taxa, traits and types, differences of degree and differences in kind. Journal of Personality, 60, 117 – 174. Meehl, P. E. (1995). Bootstraps taxometrics: solving the classification problem in psychopathology. American Psychologist, 50, 266 – 275. Meehl, P. E. (2001). Comorbidity and taxometrics. Clinical Psychology: Science and Practice, 8, 507 – 519. Meehl, P. E., & Yonce, L. J. (1994). Taxometric analysis: I. Detecting taxonicity with two quantitative indicators using means above and below a sliding cut (MAMBAC procedure). Psychological Reports, 74, 1059 – 1274. Meehl, P. E., & Yonce, L. J. (1996). Taxometric analysis: II. Detecting taxonicity using covariance of two quantitative indicators in successive intervals of a third indicator (MAXCOV procedure). Psychological Reports, 78, 1091 – 1227. Meyer, T., & Keller, F. (2001). Exploring the latent structure of the perceptual aberration, magical ideation and physical anhedonia scales in a German sample — a partial replication. Journal of Personality Disorders, 15, 521 – 535. Meyer, T., & Keller, F. (in press). Is there evidence for a latent class called ‘‘hypomanic temperament’’? Journal of Affective Disorders. Miller, M. B. (1996). Limitations of Meehl’s MAXCOV-HITMAX procedure. American Psychologist, 51, 554 – 556. Nakao, K., Gunderson, J. D., Phillips, K. A., Tanaka, N., Yorifuji, K., Takaishi, J., & Nashimura, N. (1992). Functional impairment in personality disorders. Journal of Personality Disorders, 6, 24 – 33. Oakman, J. M., & Woody, E. Z. (1996). A taxometric analysis of hypnotic susceptibility. Journal of Personality and Social Psychology, 71, 980 – 991. Raulin, M. L., Lowrie, G. S., & Brenner, V. (1994). Searching for the schizotypal taxonomy. Paper presented at the Annual Meeting of the Society for Research in Psychopathology, Coral Gables, FL. Reynolds, S. K., & Clark, L. A. (2001). Predicting dimensions of personality disorder from domains and facets of the five factor model. Journal of Personality, 69, 199 – 222. Romney, D. M., & Bynner, J. M. (1997). Evaluating a circumplex model of personality disorders with structural equation modeling. In R. Plutchik, & H. R. Conte (Eds.), Circumplex models of personality and emotions ( pp. 327 – 346). Washington, DC: American Psychological Association. Rothschild, L., Cleland, C. M., Haslam, N. & Zimmerman, M. (2002). A taxometric study of borderline personality disorder. Manuscript submitted for publication. Ruscio, J. (2000). Taxometric analysis with dichotomous indicators: the modified MAXCOV procedure and a case removal consistency test. Psychological Reports, 87, 929 – 939. Ruscio, A. M., Borkovec, T. D., & Ruscio, J. P. (2001). A taxometric investigation of the latent structure of worry. Journal of Abnormal Psychology, 110, 413 – 422.

N. Haslam / Clinical Psychology Review 23 (2003) 75–93

93

Ruscio, J., & Ruscio, A. M. (2000). Informing the continuity controversy: a taxometric analysis of depression. Journal of Abnormal Psychology, 109, 473 – 487. Sim, J. P., & Romney, D. M. (1990). The relationship between a circumplex model of interpersonal behaviors and personality disorders. Journal of Personality Disorders, 4, 329 – 341. Simpson, W. B. (1994). Borderline personality disorder: dimension or category? A maximum covariance analysis. Unpublished doctoral dissertation, Boston University. Skilling, T. A., Harris, G. T., Rice, M. T., & Quinsey, V. L. (2001). Identifying persistently antisocial offenders using the hare psychopathy checklist and the DSM-IV antisocial personality disorder criteria. Psychological Assessment, 14, 27 – 38. Skilling, T. A., Quinsey, V. L., & Craig, W. M. (2001). Evidence of a taxon underlying serious antisocial behavior in boys. Criminal Justice and Behavior, 28, 450 – 470. Soldz, S., Budman, S., Demby, A., & Merry, J. (1993). Representation of personality disorders in circumplex and five-factor space: explorations with a clinical sample. Psychological Assessment, 5, 41 – 52. Strong, D. R., Greene, R. L., & Schinka, J. A. (2000). Taxometric analysis of the MMPI-2 Infrequency scales [F and F(p)] in clinical settings. Psychological Assessment, 12, 166 – 173. Strube, M. J. (1989). Evidence for the type in type A behavior: a taxometric analysis. Journal of Personality and Social Psychology, 56, 972 – 987. Taylor, S., Rabian, B., & Fedoroff, I. C. (1999). Anxiety sensitivity: progress, prospects, and challenges. In S. Taylor (Ed.), Anxiety sensitivity: theory, research, and treatment of the fear of anxiety ( pp. 339 – 353). Mahwah, NJ: Erlbaum. Trull, T. J. (1992). DSM-III-R personality disorders and the five-factor model of personality: an empirical comparison. Journal of Abnormal Psychology, 101, 553 – 560. Trull, T. J., Widiger, T. A., & Guthrie, P. (1990). Categorical versus dimensional status of borderline personality disorder. Journal of Abnormal Psychology, 99, 40 – 48. Tyrka, A., Cannon, T. D., Haslam, N., Mednick, S. A., Schulsinger, F., Schulsinger, H., & Parnas, J. (1995). The latent structure of schizotypy: I. Premorbid indicators of a taxon of individuals at risk for schizophreniaspectrum disorders. Journal of Abnormal Psychology, 104, 173 – 183. Tyrka, A., Haslam, N., & Cannon, T. D. (1995). Detection of a longitudinally-stable taxon of individuals at risk for schizophrenia spectrum disorders. In A. Raine, T. Lencz, & S. A. Mednick (Eds.), Schizotypal personality disorder ( pp. 168 – 191). New York: Cambridge University Press. Ullrich, S., Borkenau, P., & Marneros, A. (2001). Personality disorders in offenders: categorical versus dimensional approaches. Journal of Personality Disorders, 15, 442 – 449. Waller, N. G., & Meehl, P. E. (1998). Multivariate taxometric procedures: distinguishing types from continua. Thousand Oaks, CA: Sage. Waller, N. G., Putnam, F. W., & Carlson, E. B. (1996). Types of dissociation and dissociative types: a taxometric analysis of dissociative experiences. Psychological Methods, 3, 300 – 321. Widiger, T. A., & Clark, L. A. (2000). Toward DSM-V and the classification of psychopathology. Psychological Bulletin, 126, 946 – 963. Widiger, T. A., Costa Jr., P. T. (1994). Personality and personality disorders. Journal of Abnormal Psychology, 103, 78 – 91. Widiger, T. A., & Hagemoser, S. (1997). Personality disorders and the interpersonal circumplex. In R. Plutchik, & H. R. Conte (Eds.), Circumplex models of personality and emotions ( pp. 299 – 325). Washington, DC: American Psychological Association. Widiger, T. A., & Sanderson, C. J. (1995). Assessing personality disorders. In J. N. Butcher (Ed.), Clinical personality assessment: practical approaches ( pp. 380 – 394). New York: Oxford University Press. Wiggins, J. S. (1982). Circumplex models of interpersonal behavior in clinical psychology. In P. C. Kendall, & J. N. Butcher (Eds.), Handbook of research methods in clinical psychology ( pp. 183 – 221). New York: Wiley. Woodward, S. A., Lenzenweger, M. F., Kagan, J., Snidman, N., & Arcus, D. (2000). Taxonic structure of infant reactivity: evidence from a taxometric perspective. Psychological Science, 11, 296 – 301.