Striking the (Im)Proper Balance between Scientific Advances and Clinical Utility: Commentary on the DSM-5 Proposal for Personality Disorders

Abstract

We review briefly the contributions of Skodol, Pincus, and Widiger (this issue) describing and critiquing the proposed changes in the assessment of personality and personality disorders for the DSM-5. Despite the hard work of the DSM-5 Work Group to date, there are shortcomings and areas of controversy in the current proposal that demand further attention and change. We discuss the controversy in the broader context of the DSM over the past 30 years. In addressing specific problems, we focus on the limitations of the proposed system for assessing traits (even as we endorse the movement toward dimensional assessment of personality) and the difficulties posed by the current “hybrid” model that attempts to include both traits and types. In moving forward, we suggest greater emphasis on decision-making regarding the presence and severity of any personality disorder (understood on the basis of generalized failures in adaptation) and greater flexibility in identifying the variants of personality disorders in order to accommodate both traits and types more inclusively during this transition toward dimensional approaches to assessment.

Introduction

The papers in this special section provide both an overview of the proposed changes in the assessment of personality and personality disorders (PDs) for DSM-5 (Skodol, this issue[a], this issue [b]) and critiques of these changes (Pincus, this issue; Widiger, this issue). In the first of his two papers, Skodol describes the new four-part assessment proposed, which includes identification of overall levels of personality functioning (severity of impairment), ratings of similarity to five personality disorder types, ratings of the presence of traits organized into six domains and 37 facets, and acknowledgement that the new general criteria for a PD are met. In the second paper, Skodol illustrates the use of this proposed method with three vignettes taken from the DSM-IV-TR Casebook (Spitzer, Gibbon, Skodol, Williams, & First, 2002). The new proposal generated immediate controversy during the comment period available on the DSM-5 web site, and additional critiques have now begun to appear in print (Shedler et al., 2010). The commentaries by Pincus and Widiger also voice strong reservations about the magnitude and structure of changes represented in the DSM-5 proposal.

Pincus is skeptical of the validity of prototype ratings (for both functional impairment and PD types) and the deletion of five PDs from the nosology, with particular concern about the loss of narcissistic PD. He is supportive, however, of the “two-step diagnostic process” now proposed, that is, a process that distinguishes general diagnostic criteria for any PD (presence of the PD “genus”) from identification of the specific variant of PD (description of the PD “species”). We appreciate his genus/species analogy and agree that the critical, first decision should be that regarding the presence versus absence of a PD. We are concerned that many clinicians will simply default on making decisions regarding PDs given the complexity of the current proposal, and we are eager to encourage a more accessible process in which decisions about the presence (“genus”) and severity of a PD are primary, with finer distinctions made possible by additional ratings of personality traits and types. Distinguishing among the particular variants (“species”) of personality dysfunction is necessarily more complex (whether using traits, prototypes, criteria sets, or Q-sort items), and if clinicians undertake this, they will require better decision support and psychometric tools than are available under the current proposal, a topic we discuss below.

Widiger organizes his comments around the general theme that the current proposal lacks sufficient empirical grounding (in virtually all its aspects) and then focuses on three specific issues: the deletion of five of 10 of the current Axis II diagnoses, the abandonment of sets of diagnostic criteria for prototype matching, and the choice of the dimensional model of traits incorporated into the new proposal. Widiger has been a longstanding and outspoken advocate of the use of the Five-Factor Model (FFM) of personality as a potential system for PD diagnosis, and he has also proposed a four-step methodology for using the FFM to organize diagnostic decision-making for PDs (Widiger & Mullins-Sweatt, 2009). In a paper integrating the FFM (as operationalized in the NEO-PI-R, Costa & McCrae, 1992) with the dimensional models assessed by the Dimensional Assessment of Personality Pathology—Basic Questionnaire (DAPP-BQ; Livesley & Jackson, in press) and the Schedule for Nonadaptive and Adaptive Personality (SNAP-2; Clark, Simms, Wu, & Cassillas, in press), Widiger, Livesley, and Clark (2009) have also offered specific suggestions for how such an integrated dimensional approach could be used for diagnostic purposes. We agree with Widiger on many of the problems inherent in the prototype matching system and the personality trait structure now proposed.

As Pincus and Widiger have noted, the changes proposed by Skodol qualify as “substantial” or “major” by the criteria developed by the DSM-5 Task Force (Kendler, Kupfer, Narrow, Phillips, & Fawcett, 2009). Such changes require “a broad consensus of expert clinical opinion” (p. 5), but thus far, disagreement has predominated (Shedler et al., 2010). Such changes also require substantial empirical support, as demonstrated by diagnostic stability, response to treatment, and predictive validity (Kendler et al., 2009), and few commentators (including Pincus and Widiger) have been persuaded on this basis. We expect that arguments marshalling competing (and often selective) evidence in support of a particular diagnostic approach for DSM-5 will continue to appear (an important step in sharpening the positions involved), but this process itself will serve to emphasize the lack of research and clinical consensus.

The Broader Context of the DSM

It is useful to consider the current proposal in the broader historical context of the DSM. This context includes the important role of the DSM in identifying and characterizing personality disorders over the past 30 years, the inevitable tension that exists in the DSM because of its dual role as a clinical document and a research document, and the significant advances that have been made in the field of personality psychology during the past several decades. On this basis, it will be possible to assess more clearly the coherence and impact of the current proposal for DSM-5 and the criticism that it has received.

Supporting the Legitimacy of Personality Disorders

The most important contribution of the DSM since 1980 (with the publication of DSM-III and the emergence of a separate Axis II for personality disorders) has been to increase the legitimacy of personality disorders as both formal diagnoses and worthwhile topics for research investigation. This assertion may sound like a truism to those immersed full-time in the classification, study, and treatment of PDs, but even today, we should not take for granted the credibility of these disorders in the eyes of some mental health experts, general medical practitioners, and the lay public (see Tyrer, 2008). PDs have a substantial prevalence, especially in clinical settings, both medical and psychiatric (Tyrer, 2008); they are associated not only with greater utilization of treatment (Bender et al., 2001) but also with a more guarded prognosis (Grilo et al., 2005; Newton-Howes, Tyrer, & Johnson, 2006); and they are linked to chronic impairments in functioning (Skodol et al., 2002; Seivewright, Tyrer, & Johnson, 2004), the hallmark of PDs (Rutter, 1987). Given the public health significance of PDs, any new proposals should not threaten their legitimacy (“first, do no harm”), not only for mental health professionals but also for the broader audience of clinicians (Kupfer & Regier, 2010) and the interested lay public to whom the DSM is intended to speak. Unfortunately the controversies surrounding the current DSM-5 proposal present a public picture that is confusing for the outside observer, either clinical or lay.

The lack of continuity with previous versions of the DSM is pronounced and risks undoing the progress of the past 30 years rather than being generative as intended. Clinicians and researchers familiar with the 10 PDs defined in DSM-IV are likely to be jarred by the proposal, which emphasizes a new trait rating system, fundamentally alters the definitions of 5 PDs, and deletes the remaining 5 PDs, aggregating them into a single category with the unwieldy title of “Personality Disorder Trait Specified.” In this vein, it is surprising that there is no discussion by Skodol of “possible unintended negative effects” (Kendler et al., 2009, p. 2) of the proposed changes. We believe that such negative effects, even if unintended, are especially likely to undermine the clinical utility and application of the proposed changes. If providers choose to “defer” PD diagnoses (a concern also expressed by Widiger) because the proposed ratings are not familiar, insufficient rationale is provided, or the decision-making process is too daunting, then patients suffering from PDs will go unrecognized and untreated, and the negative impact of PDs on the course of chronic health conditions (both mental and physical) will persist (Gross et al., 2002; Newton-Howes et al., 2010).

The Tension between Clinical Utility and Research Grounding

The DSM must serve two (sometimes competing) purposes. On the one hand, the DSM must provide a comprehensive classification system that leads to valid and clinically informative diagnoses in mental health treatment settings: “The DSM is above all a manual to be used by clinicians…” (Kendler et al., 2009, p. 1). On the other hand, the DSM is intended to reflect the current state of empirical research and to point to productive ways forward for improving our understanding of the mechanisms related to the development and maintenance of PDs: “Recommendations should be guided by research evidence.” (Kendler et al., 2009, p. 1) The goal of such understanding is to promote the development of more effective approaches to prevention and treatment. Unfortunately, these different purposes often aggravate (rather than soften) potential divisions between clinicians and scientists, and any proposal must do an effective job of accommodating the needs of both constituencies. The goal should be to use the clinician/scientist tension in creative ways, e.g., by tempering complex research agendas with an awareness of the pragmatic needs of clinicians, while, at the same time, informing clinicians about “best practices” for classification, assessment, and diagnosis based on the most compelling research evidence. We believe that the current proposal is not sensitive enough to this (admittedly difficult) balancing act and fails to provide a comprehensive integration that is suitable for both needs—there is too much complexity and too little continuity for clinical purposes and too little systematic, empirical corroboration for research purposes. On balance, the “voice” of the clinician appears to be particularly lacking, although one of the goals of the DSM field trials will to be to solicit feedback from the clinical arena (Kraemer, Kupfer, Narrow, Clarke, & Regier, 2010).

The Relationship between Models of Normal Personality and Personality Disorders

A notable achievement over the past several decades in personality psychology has been an evolving understanding of how normative models of temperament and personality can be used to characterize the variants of personality disorders (Clark, 2007; Widiger & Lowe, 2008), contributing to a synthesis of work across these areas, a fact that is reflected in the title of the DSM-5 Personality and Personality Disorders Work Group itself (emphasis added). Two related themes are prominent in research on the classification of PDs from the past 20 years: the superiority of dimensional models of personality disorders (and psychopathology more generally) over existing categorical models in terms of predictive validity, comorbidity, and coverage of key personality features (Widiger & Trull, 2007); and the alignment between models of abnormal and normal personality (Markon, Krueger, & Watson, 2005).

There is now strong evidence that dimensional models of PDs have many psychometric and conceptual advantages over the categorical system represented in the DSM and that models of normative personality are largely able to represent constructs relevant to abnormal personality (Clark, 2007). Thus, the emphasis of the DSM-5 proposal on personality traits as a basis for PD diagnosis is well-grounded, and we applaud the intellectual spirit of this revision. There is a need to incorporate a sound structural model of personality into the assessment of PDs so that the dialogue among research evidence, clinical utility, and modern psychometric methods can continue in productive ways. One major concern, however, is the limitations in the trait model currently proposed, the topic to which we now turn.

How Best to Incorporate the Assessment of Traits?

The current proposal suggests a trait model with six domains (and 37 facets). The six domains reflect the extreme ends (maladaptive variants) of Negative Emotionality, Detachment, Antagonism, Disinhibition, Compulsivity, and Schizotypy. We are concerned that this coverage fails to provide a comprehensive picture of personality traits, which limits the validity and utility of this classification system. The proposed trait model reflects an attempt to integrate separate, competing personality models into an inclusive trait rating system based on the best available science, an effort we endorse. Skodol argues that this proposed system is a good representation of the four-factor model of personality (Widiger & Simonsen, 2005; Widiger, Livesley, & Clark, 2009) inherent in many competing models (e.g., FFM, DAPP-BQ, and SNAP-2), supplemented by dimensions for compulsivity (in order to capture more of the dysfunctional features associated with obsessive-compulsive PD) and schizotypy (to provide enhanced coverage of the aberrant cognitive experiences characteristic of Cluster A PDs).

We agree with the principles guiding the development of the proposed trait rating system. However, we believe that the current model does not adequately reflect the four-factor (plus oddity/openness) structure that has been empirically supported. The four-factor model found to represent a common structure includes four bipolar domains, with the assumption that exaggerated expressions at both ends of the bipolar continua are problematic: (1) heightened negative emotionality (neuroticism) versus insensitivity to negative affect (maladaptive forms of fearlessness and indifference to negative emotional states), (2) extraversion and positive emotionality versus introversion and detachment, (3) agreeableness versus antagonism, and (4) conscientiousness and constraint versus impulsivity and undependability (Widiger, Livesley, & Clark, 2009). Watson, Clark, and Chmielewski (2008) found evidence for an oddity dimension characterized by unusual perceptual and cognitive experiences that adds distinctive information beyond more normative expressions of openness to experience, and they suggested that oddity be incorporated into any system relevant to personality disorders.

From a psychometric perspective, the use of unidimensional constructs rated on a unipolar scale is both a conceptual advantage—the construct is easier to understand—and a practical advantage—the usual principle is to try to measure only one thing at a time and to create the best possible “ruler” for that construct across the full range of severity (from absent to strongly present). The use of models from item response theory (IRT), which rely on an assumption of unidimensionality, has many desirable features in this regard, and such models are now being used increasingly for applications in personality and psychopathology (Reise & Waller, 2009). From this perspective, the bipolarity of the FFM domains (as argued by Widiger and others) is less desirable, and a system incorporating unipolar ratings requires 10 dimensions: (1) negative emotionality, (2) indifference to negative emotion, (3) extraversion, (4) introversion, (5) agreeableness, (6) antagonism, (7) constraint, (8) disinhibition, (9) oddity, and (10) conventionality and concreteness. Only six of these dimensions are represented in the current proposal, and the constraints on this dimensional space have resulted in some of the lower-order facets appearing to be out of place (e.g., the “histrionism” facet is listed under the antagonism domain; see Widiger, this issue).

We are aware of the added complexity that the use of more trait domains would create, and such complexity makes even more pressing the need for systematic tools to help operationalize such a process and to interpret its results. To make such a process manageable, it could be applied adaptively, a suggestion also made by Widiger, Livesley, & Clark (2009). That is, clinicians could rate the 10 higher-level domains as a first step, with further ratings at the facet level only if a specified threshold is surpassed on the domain rating (the current proposal also acknowledges this as an option). We have mixed feelings, however, about the use of ratings only at the domain level. First, we recognize the general finding that facets often add more information than domains alone (cf. Samuel & Widiger, 2008). Second, facets conform more closely than domains to the themes and criteria familiar to clinicians when characterizing PDs and identifying priorities for treatment. In any case, “psychoeducation” and training needs to be provided to clinical audiences about the improved diagnostic formulation that follows from systematic and disciplined attention to the trait profiles of patients. It may be small comfort, but the use of 10 domains (and associated facets) is still likely to involve fewer (but more inclusive) indicators that the current 79 criteria (together with the 15 markers for conduct disorder) used to define the 10 DSM-IV PDs.

Serious concerns have been raised about the openness of clinicians to complex trait-based rating systems relative to prototype matching (First, 2005; Rottman, Ahn, Sanislow, & Kim, 2009). One way to ameliorate these concerns would be to provide validated instruments for the assessment of DSM-5 PD traits, together with norms, scoring procedures, and interpretation guidelines. In the current proposal, brief descriptions of each facet are provided, but many of these constructs are complex and inferential, opening the door to idiosyncratic interpretation by diagnosticians and unclear decision procedures for determining the level of each trait (absent to extremely descriptive). Virtually all models of normal and abnormal personality have been developed and validated by constructing self-report, interview, or clinical rating instruments and subjecting these to rigorous psychometric testing, and the DSM-5 work group would do well to develop such assessment tools for their proposed taxonomy. Providing such instruments would ensure that clinicians assess DSM-5 personality constructs in similar ways and receive clear feedback about trait profiles (e.g., knowing that a patient ranks in the 95^th percentile of negative emotionality relative to the general population is more informative that a clinician inferring that the patient has “moderate” or even “extreme” negative emotionality).

Is It Possible to Accommodate Both Traits and Types?

We recognize both the heuristic value of prototypes (Shedler et al., 2010), especially for clinicians educated in a context of categorical diagnosis, and the intuitive appeal for some audiences of person-centered classification (types or classes of people) versus variable-centered approaches (a multidimensional space of traits or factors). Nonetheless, the value of the prototypes proposed in the current DSM-5 proposal is problematic. Under the proposed system, clinicians would rate patients in terms of their overall fit to each narrative and provide ratings of several traits presumed to characterize each type. The types nominally represent distinctive constellations of personality traits that have clinical implications, such as prognosis or treatment selection, and several types (especially borderline and antisocial) build upon a large research literature documenting the construct validity of these syndromes. Each type, however, is defined by an uneasy combination of two components: a short narrative describing the type, and a set of traits (drawn from the more inclusive trait rating system) that best characterize the type.

In this prototype matching system, clinicians are asked to rate their impression of how well the narrative characterizes a given patient, and the anchors for different degrees of match are general. Although there is some empirical support for this approach to PD diagnosis (Westen, Shedler, & Bradley, 2006), further validation is needed, particularly given problems with interrater reliability that stimulated the transition from the prototype-based DSM-II to the criterion-based DSM-III. Under the new proposal, diagnosticians are given carte blanche to interpret each prototype narrative in potentially different ways, opening the door to a host of known problems with cognitive heuristics, such as salience and availability biases (Kahneman, Slovic, & Tversky, 1982). It is difficult to conceive of how such an unsystematic approach will enhance the construct validity of PDs.

The presence of these two definitional components—a narrative descriptive with many embedded features and a select group of traits—gives rise to several questions. Why rate the traits when one has rated the prototype, or vice versa? Could one derive a prototype similarity score simply by correlating the trait scores with a “gold standard” prototype profile? Such an approach has been used with the FFM (Lynam & Widiger, 2001; Miller, Bagby, Pilkonis, Reynolds, & Lynam, 2005) and is likely to be more reliable psychometrically (see also Pincus, this issue.) More broadly, if types are fundamentally constellations of traits, why retain types at all when trait ratings can be completed exhaustively and more precisely? Indeed, rating traits for each type, then rating the same traits in the third phase of Skodol’s diagnostic progression seems redundant.

Skodol argues (this issue[a]) that personality types should be a component of the new diagnostic system because of evidence that dimensional and categorical models of personality account for important variability in clinical outcomes not accounted for by the other model. Whereas the Work Group appears to have interpreted the incremental validity literature as evidence in favor of a trait-plus-type approach to PD classification, we view this issue differently. Research on incremental validity in personality does not give special preference to categorical models per se, but finds instead that many models—continuous and categorical—overlap extensively in their concurrent and predictive validity and that each model also explains a unique proportion of variance in clinical outcomes. A common finding, however, is that the proportion of variance uniquely accounted for by any one model over another is quite small. As a result, the best approach is likely to be the development of a single, integrated model (Widiger & Simonsen, 2005) rather than the hybrid model now proposed.

The critiques provided by Pincus and Widiger also highlight the insufficient rationale provided by the DSM-5 proposal for eliminating five of the DSM-IV PD diagnoses, while restructuring and reformulating the remaining five. Pincus elaborates on the significant shifts in the content of the DSM-5 types relative to the DSM-IV diagnoses by focusing on the elimination of narcissistic PD. Narcissistic grandiosity, for example, has become part of the antisocial type, despite little evidence for this shift, and substantial evidence for the validity of narcissism (and its two variants—grandiose and vulnerable) as a key construct. In this vein, it is ironic that Skodol (this issue[a]) points to “narcissistic difficulties” as a “core dimension” across personality psychopathology.

We have examined the proposed prototype for borderline PD and compared the narrative text to the criteria in DSM-III-R and DSM-IV (see Table 1). The table demonstrates that the content in the DSM-5 type is significantly different from that in the DSM-IV criteria set and that the change between DSM-IV and DSM-5 is much larger than that between DSM-III-R and DSM-IV. In addition, the DSM-5 narrative description is broader and more general than the specific criteria in previous versions of the DSM, opening the door to greater variability in interpretation. It also contains language hypothesizing specific mechanisms or causes of behavior. While we endorse the importance of interpersonal mechanisms in the etiology and maintenance of BPD (Choi-Kain, Zanarini, Frankenburg, Fitzmaurice, & Reich, 2010; Hill et al., 2008; Morse et al., 2009), other models have been articulated and supported by empirical work (e.g., Beck & Freeman, 2004; Linehan, 1993). Clinicians or researchers who endorse other models may be less likely to rate the borderline PD type in consistent ways because they interpret the same behavior differently.

Table 1.

A Comparison of the Criteria for Borderline Personality Disorder

DSM-III-R Diagnosis		DSM-IV Diagnosis	DSM-5 Prototype Narrative	DSM-5 Trait Facet (Domain)
A pervasive pattern of instability of mood, interpersonal relationships, and self-image, beginning by early adulthood and present in a variety of contexts, as indicated by at least FIVE of the following:		A pervasive pattern of instability of interpersonal relationships, self-image, and affects, and marked impulsivity beginning by early adulthood and present in a variety of contexts, as indicated by FIVE (or more) of the following:
Criterion number 1	a pattern of unstable and intense interpersonal relationships characterized by alternating between extremes of dealization and devaluation	Criterion number 2	As a result [of extremely fragile self-concept], they have an impoverished and/or unstable self structure and difficulty maintaining enduring intimate relationships. Thus, interpersonal relationships are highly unstable and alternate between excessive dependency and flight from involvement. Empathy for others is severely impaired.	Separation insecurity (NE)
2	impulsivity in at least two areas that are potentially self-damaging, e.g., spending, sex, substance use, shoplifting, reckless driving, binge eating. Do not include suicidal or self-mutilating behavior covered in Criterion 5	4	Individuals with this type are characteristically impulsive, acting on the spur of the moment, and frequently engage in activities with potentially negative consequences. Intense distress may also lead to other risky behaviors, including substance misuse, reckless driving, binge eating, or promiscuous sex.	Impulsivity (DS)
3	affective instability [due to a marked reactivity of mood] [(e.g., intense episodic dysphoria, irritability, or anxiety usually lasting a few hours and only rarely more than a few days)]	6	Individuals with this disorder experience rapidly changing, intense, unpredictable, and reactive emotions and can become extremely anxious or depressed. Emotional reactions are typically in response to negative interpersonal events involving loss or disappointment.	Emotional lability (NE) Anxiousness (NE) Depressivity (NE)
4	inappropriate, intense anger or lack of control of anger, [(e.g., frequent displays of temper, constant anger, recurrent physical fights)]	8	They may also become angry or hostile, and feel misunderstood, mistreated, or victimized. They may engage in verbal or physical acts of aggression when angry.	Hostility (A) Aggression (A)
5	recurrent suicidal threats, gestures, or behavior, or self-mutilating behavior	5	Deliberate acts of self-harm (e.g., cutting, burning), suicidal ideation, and suicide attempts typically occur in the context of intense distress and dysphoria, particularly in the context of feelings of abandonment when an important relationship is disrupted.	Self-harm (NE)
6	[identity disturbance: markedly and persistently unstable self-image or sense of self]	3	Individuals who match this personality disorder type have an extremely fragile self-concept that is easily disrupted and fragmented under stress. At the same time, intense, intimate involvement with another person often leads to a fear of loss of an identity as an individual.
7	chronic feelings of emptiness	7	[Fragile self-concept] results in the experience of a lack of identity or chronic feelings of emptiness.
8	frantic efforts to avoid real or imagined abandonment. (Do not include suicidal or self-mutilating behavior covered in Criterion 5)	1	Relationships are based on the fantasy of the need for others for survival, excessive dependency, and a fear of rejection and/or abandonment. Dependency involves both insecure attachment, expressed as difficulty tolerating aloneness; intense fear of loss, abandonment, or rejection by significant others; and urgent need for contact with significant others when stressed or distressed, accompanied sometimes by highly submissive, subservient behavior.	Separation insecurity (NE)
---	Transient, stress-related paranoid ideation or severe dissociative symptoms	9	Quasi-psychotic reactions, including paranoia and dissociation, may progress to transient psychosis.	Dissociation proneness (S)
			Self-appraisal is often associated with self-loathing, rage, and despondency.	Low self-esteem (NE)
			Core emotional traits and interpersonal behaviors may be associated with cognitive dysregulation, i.e., cognitive functions may become impaired at times of interpersonal stress leading to information processing in a concrete, black-and white, all-or-nothing manner.

Finally, as noted by Pincus, the ontological status of PD types is a separate issue from decisions regarding whether or not to include certain types in the DSM and how to operationalize them. Our view is that dimensional and categorical models of the latent structure of PDs should be compared directly using information theoretic approaches (Hallquist & Pilkonis, 2010; Markon & Krueger, 2006). Moreover, hybrid models of PDs that incorporate both categorical types and continuous dimensions may, in fact, prove to best represent variation in personality and psychopathology (Muthén, 2006). The fitting of such models should be done empirically, however, and a potential contribution of hybrid analyses of the personality trait system could be to identify personality types in a data-based way, rather than carrying forward (or eliminating) constructs from previous editions of the DSM on the basis of more subjective judgments of the quality of existing research evidence. For example, if borderline PD is adequately represented by traits in the new system and is a distinctive category, one would expect a latent class analysis of the traits to extract a borderline subgroup. Failure to identify such a subtype could suggest problems with the coverage of the trait system or could reflect problems in the BPD construct.

Translating from Traits to Types

One primary intention of the DSM-5 proposal is to promote the use of a dimensional approach to assessment and diagnosis that treats personality as a multivariate profile of traits (Clark, 2007). Given this stance, however, we believe that the proposal should include explicit provisions to help mental health professionals make the transition to the new system—the shift in emphasis from categorical to dimensional classification of PDs will require considerable time to achieve wide acceptance, perhaps occurring only with multiple iterations of the DSM system. Therefore, despite our many reservations about the current format for including types, we believe that there is a persuasive argument for the retention of types. However, it is a pragmatic argument rather than a conceptual one—that is, to facilitate the major transition from categories to dimensions. In this light, it would be valuable to retain all the categorical prototypes proposed since DSM-III rather than supplanting existing categories with the new personality types system and to provide the rationale and tools for effective translations from traits to types. The justification for many earlier types (including provisional ones) was often made on clinical grounds and acknowledging the potential value of such previous contributions for this new movement toward dimensional models can serve to integrate past and present more comprehensively and gracefully. In this regard, greater inclusiveness is likely to be a virtue that leads to greater eventual consensus—see, for example, Blagov, Bradley, and Westen (2007) on the “subthreshold” personality types that emerge with the use of the Shedler-Westen Assessment Procedure-200 (SWAP-200, Westen & Shedler, 1999) and that, in principle, could also be mapped using a comprehensive set of traits.

To help professionals understand how PD categories align with the new trait system, detailed translation tables from dimensional traits to categorical PDs could be provided. Currently there are no algorithms for determining types on the basis of traits, although the work with the FFM mentioned above provides one model. One long-term hope is that the wisdom of using a comprehensive trait-based approach to assessment will become apparent when the flexibility of such a system (especially at the level of facets) is clear. Models like the FFM are grounded in latent constructs that have a broad reach and sound empirical support because of their origins in natural language. In turn, they can be used to “map” the outcomes that result with the use of other criteria sets (e.g., DSM diagnostic features, Q-sort items from the SWAP), making core traits an effective “lingua franca” for characterizing both personality and personality disorders—and it is, after all, a shared language that we are hoping to create with any diagnostic system. Prototypes are evocative for clinicians, but it is usually the psychological themes and mechanisms represented by the weaving together of key traits (e.g., affective instability, reflected in a combination of vulnerability, impulsiveness, and angry hostility, all facets of the FFM) that define the lived experience of our patients and provide critical targets for treatment.

In addition, structural models of personality capture important aspects of temperament (Clark, 2005) and expressive style (e.g., internalizing versus externalizing; Kendler et al., 2010) that are likely to be linked more directly to underlying genetic and biological substrates of behavior than the behavioral features of PD phenotypes. The National Institute of Mental Health has launched a Research Domain Criteria (RDoC) initiative (Insel et al., 2010) that is attempting to identify transdiagnostic constructs that have explanatory power for many forms of psychopathology and that can be linked to underlying mechanisms of pathophysiology. Work on personality disorders, anchored in the best current models of personality and temperament, is compatible with such an approach.

Implications for a New Paradigm in DSM-5

We believe that our comments lead naturally to an assessment process that differs from the current proposal. The first (and most critical) decision should be one about the presence versus absence of any PD (i.e., Pincus’ “genus” decision) based on a definition that emphasizes persistent and pervasive failures in general adaptation (Livesley, 1998). Such a decision would represent a combination of steps 1 and 4 in the current proposal: the decision that the patient does, indeed, meet the general criteria for a PD (which currently is left for last) and a rating of PD severity.

Despite our enthusiasm for using a trait-based approach to identifying the specific variants (or “species”) of PD, we do not believe that PDs are only the result of extreme manifestations of traits. Such exaggerated traits (well captured by conventional models of personality) are risk factors, but they are unlikely to be sufficient explanations for the development of PDs, which are characterized fundamentally by impairments in functioning and adaptation Problematic variants in temperament and personality may complicate the development of adaptive competencies, but they do not ensure failure—such failures are likely to be linked as well to additional forms of developmental adversity. Morey and Zanarini (2000), for example, provided a cogent example of how a history of early adversity adds incremental validity beyond that attributable to FFM traits in the characterization of borderline PD.

We believe, however, that the current formulation of the general criteria (and the specific guidance for rating the severity of “impaired sense of self-identity” and the “failure to develop effective interpersonal functioning”) are too esoteric and inferential for general clinical use and too narrow theoretically to accommodate the diversity of research agendas reflected in the area of PDs. We believe that a more transparent definition and better manifest indicators of these latent constructs need to be developed. For guidance, we point to two useful sources: first, the circumplex model (see Pincus, this issue), which is the model of interpersonal functioning that has the strongest grounding in basic behavioral research on interpersonal relatedness, and second, discussions of the important developmental tasks confronted in adulthood. For the latter, the classic schema remains that of Erikson (1963), but important related thinking is represented, for example, in Bowlby’s work (1979) on the role of attachment in adulthood (and the necessary balance between attachment and autonomy) and Blatt’s work (2008) on the “polarities of experience” (and the dialectic between relatedness and self-definition). In the first case, the circumplex model is organized around themes of agency and communion (“to work and to love”). In the second case, the developmental tasks of adulthood can be characterized as the need to establish a sense of autonomy, achievement, and self-definition (identity vs. confusion); the capacity for attachment, reflected in both intimate relationships and indicators of broader social integration (intimacy vs. isolation); and the capacity for generativity and prosocial engagement on behalf of others and the “social contact” (generativity vs. self-absorption).

These concepts (and this language) are not radically different from the current proposal (e.g., Skodol, this issue[a], cites these and related sources), but the proposal does use a vocabulary drawn more exclusively from object relations and social cognitive theories that suggests a narrower focus and evokes connotations that may be polarizing for some audiences. A more “neutral” and developmentally oriented language is likely to be more broadly acceptable and to point to manifest indicators of failure in adaptation that are more transparent (e.g., obvious failures in role performance) and easier to rate reliably. Such indicators are likely to be correlated with more inferential indicators and to provide comparable information. In this vein, Mintz (1981) provided a classic example of the convergence between relatively simple ratings of symptomatic improvement and more complex and labor-intensive judgments of psychodynamic changes in psychotherapy. An earlier discussion of potential general diagnostic criteria for a PD by Krueger, Skodol, Livesley, Shrout, and Huang (2007) is also more consistent with these suggestions than the current proposal.

To expand further, we suggest consideration of revised general criteria that would capture the three themes mentioned above. The first is a failure to achieve sufficient autonomy and self-definition reflected in the:

• Inability to achieve in the areas of education, work, and task-oriented accomplishment.

• Inability to develop a consistent and realistically balanced representation of oneself. We believe that this feature probably remains the most elusive marker to rate and that more work, in addition to the current attempts by the DSM-5 Work Group, needs to be done on operationalizing such a criterion satisfactorily.

The second theme is failures in interpersonal relatedness, both in terms of close personal ties and broader social integration. These are reflected in the:

• Inability to develop and maintain supportive attachment figures (this is typically reflected in romantic relationships in our culture, but may also be relevant to other committed relationships, e.g., close friends, first-degree relatives, or other family ties).

• Inability to develop and maintain more general social integration (e.g., friendships, involvement in community activities and affairs). Weiss (1998) characterized nicely the need for both close relationships and more general social integration (“affiliations”) and the different provisions that these ensure. Hill, Pilkonis, and Bear (2010) also discussed the importance of different social domains and a developmental approach to PDs based on evolving competencies across domains.

The third theme is a failure in generativity, reflected in the:

• Inability to engage with a purpose beyond one’s own self-interest, goals on behalf of succeeding generations, or other larger goals. In other terms, this is often reflected in an inability to develop a sense of meaning in life and in a phenomenology characterized by themes of estrangement and alienation.

• Imposition of distress on others. The person with the PD is not always the person who suffers the most as a consequence, and any general criteria for the diagnosis of PD should include markers for the social burden associated with PDs.

Consistent with the intentions of the current proposal, ratings of the severity of these indicators should be developed, but given our comments above on the psychometric utility of unidimensional (and unipolar) ratings, separate ratings of severity of the different components should be done. Such separate ratings could then be used to inform a global rating of severity and a diagnostic decision about the presence versus absence of a PD. In principle, it would be possible to model (using regression approaches) the elements that are most influential in making the global severity and yes-versus-no PD decisions and to provide more explicit rules and thresholds for coming to such a decision, especially as data from large clinical samples are collected.

After the presence of the PD “genus” has been thoughtfully established, we believe that a critical choice should be offered to the diagnostician—a choice to stop after an attempt to derive a valid decision about the presence and severity of a PD (given limitations of time, resources, or interest) or to continue with the next step in the process, which would involve identifying specific variants of PDs (at an additional cost of time and effort). We believe it is preferable to stop with careful yes/no and severity decisions rather than to insist on a “shallow” or noisy attempt to rate further indicators.¹ As mentioned above, the presence of any PD has important prognostic value, and it would be better to have such general information available to guide decisions about the type, level, and intensity of treatment resources required rather than to default to deferring a diagnosis.

If the choice is made to proceed, then step two in the process should be a review of a comprehensive trait profile in order to ensure that all relevant personality themes are represented in the diagnostic formulation. As mentioned above, the clinician should have a set of psychometric tools and adequate decision support for this purpose (e.g., a dictionary of definitions of trait domains and facets, structured guides and score sheets, realistic expectations about the time and effort involved to perform the task adequately). For the sake of efficiency, the trait-rating task could be done adaptively, i.e., the use of ratings that begin with domains and then proceed to facets only if the domain is relevant, recognizing, however, that the use of facets leads to a richer clinical portrait.

Finally, the third step in the decision-making process could be undertaken, that is, to organize the specific trait-based ratings into prototypes for the sake of cognitive efficiency, consistency with historical precedents, and bridge-building during the transitional from categorical to dimensional approaches to assessment. We believe that prototype ratings should be saved for last in order to avoid the pitfalls of misleading cognitive heuristics, i.e., the possibility of clinicians to jumping too quickly to conclusions about types on the basis of salience or availability biases influenced by a few highly visible, selective, or state-related markers. In this scenario, there is no reason to limit the number of prototypes that could be noted. All relevant ones could be acknowledged, and in the service of differential diagnosis, they could be rank ordered or rated for goodness-of-fit, consistent with the current proposal.

One long-term psychometric goal would be to develop better rule-based methods for translating from traits to prototypes. Initial algorithms could be provided, based on previous work in this vein (Lynam & Widiger, 2001; Miller et al., 2005), and the collection of additional systematic data on both traits and prototypes would further promote this effort, even as we move toward a more dimensionally based system. Despite his best efforts to justify the current proposal, Skodol himself (this issue[b]) supports the utility of previous types in his discussion of case vignette #2, “Dr. Wilson.” He provides the best summary statement about Dr. Wilson when he describes him succinctly as “a textbook case of passive-aggressive personality disorder, with narcissistic features,” but then must devote himself to a lengthy explanation of why Dr. Wilson would now be classified as a moderate match to the new antisocial/psychopathic type and a slight match to the new obsessive-compulsive type, which evokes a very different clinical picture for seemingly arbitrary reasons.

Concluding Comments

Despite the hard work of the DSM-5 Work Group to date, there are shortcomings and areas of controversy in the current proposal that demand further attention and change. Pincus, Widiger, and others have provided critiques that range broadly, but in general terms, the criticisms are organized around issues of consensus, complexity, continuity, content, and corroboration, five “C’s” that require correction if the DSM is to be effective in its fifth edition. Thus, the changes proposed are major and do not reflect a consensus in the field at present; the multi-stage, “hybrid” model proposed introduces layers of complexity that are likely to be too daunting for usual diagnostic purposes; there is a lack of continuity with previous editions of the DSM that may threaten the legitimacy of PDs as important diagnoses and topics of investigation and that will be jarring to clinicians; there are major shifts in content in traits and prototypes relative to the empirical literature and previous editions of the DSM; and there is a lack of systematic empirical corroboration for many parts of the proposal. In addition to identifying these problems, we have offered some suggestions for addressing them, and we look forward to the continued dialogue that will surround the future work of the DSM-5 Work Group.