Introduction

Lanier, Bollinger, and Krueger (2011) provide an overview of proposed changes to the diagnosis and classification of personality disorders to appear in the forthcoming fifth edition of the American Psychiatric Association’s (APA) Diagnostic and Statistical Manual of Mental Disorders (DSM-5). They are correct that the proposed changes to the personality disorders section are fundamental, and in some respects may represent a true paradigm shift in how a mental disorder is conceptualized and diagnosed. As expressed by the Chair of the DSM-5 Personality and Personality Disorders Work Group (PPDWG), “the work group recommends a major reconceptualization of personality psychopathology” (Skodol 2010, “Reformulation of personality disorders in DSM-5,” para. 1). I do not myself disagree with some of the more radical proposals that are being made, but a difficulty I do have is the surprisingly liberal threshold that has been accepted for making any such revisions (Frances 2009; Widiger 2011). In addition, one point of strong agreement with Lanier et al. is that the construction of DSM-5 is in a state of “flux,” to the point that it is frankly difficult to predict or anticipate what may in fact happen to the diagnosis of personality disorders.

Lanier et al. state that the proposals for DSM-5 consist of four major changes (see Table 1 of Lanier et al.): (1) a change to the definition of personality disorder to require the presence of a pathology of identity integration, integrity of self-concept, and self-directedness in order for a diagnosis to be made; (2) an assessment of level of self and interpersonal dysfunction; (3) the deletion of five diagnoses; and (4) the inclusion of a six domain (37 trait) dimensional trait model. However, since this paper was written, there have been quite a few significant changes to these (and other) proposals.

Prototype Matching

Missing from the Lanier et al. list of changes was a decision by the PPDWG to abandon diagnostic criterion sets for prototype matching (Skodol 2010; Skodol et al. 2011). One of the, if not the, major innovation of the third edition of the APA diagnostic manual (i.e., DSM-III; APA 1980) was a shift away from the unreliable prototype matching to the requirement that a mental disorder diagnosis include a systematic and comprehensive assessment of a specific and explicit set of diagnostic criteria (Spitzer et al. 1980). The criterion sets of DSM-III increased dramatically the ability of researchers to conduct reliable, replicable, and valid research. As expressed recently by Kendler et al. (2010), “the renewed interest in diagnostic reliability in the early 1970s-substantially influenced by the Feighner criteria- proved to be a critical corrective and was instrumental in the renaissance of psychiatric research witnessed in the subsequent decades” (p. 141). One of the benefits of this renaissance was the highly published Collaborative Longitudinal Studies of Personality Disorders (CLPS), which used as its primary measure a semi-structured interview that systematically assessed the DSM-IV personality disorders’ specific and explicit criterion sets (Skodol et al. 2005).

Nevertheless, the PPDWG proposed to abandon diagnostic criterion sets for prototype matching, in which one matches one’s perception of a patient with a 10–17 sentence paragraph description of a prototypic case (Skodol 2010; Westen et al. 2006). “To make a diagnosis, diagnosticians rate the overall similarity or ‘match’ between a patient and the prototype using a 5-point rating scale, considering the prototype as a whole rather than counting individual symptoms” (Westen et al. 2006, p. 847). Rather than require a researcher or a clinician to spend 2–4 h carefully assessing each diagnostic criterion, with prototype matching “clinicians could make a complete Axis II diagnosis in 1 or 2 min” (Westen et al. p. 855) because one does not assess each individual sentence within the narrative description. Instead, the clinician matches their perception of the patient with the overall gestalt. The diagnosis is reduced from a systematic assessment of each of the nine diagnostic criteria for DSM-IV-TR borderline personality disorder (or each of the 13 sentences within the DSM-5 narrative description of a prototypic case) to simply a single judgment: whether the patient’s personality appears to match the set of sentences, considered together as a unified whole.

This proposal was made despite the fact that there is a considerable body of research to document the poor reliability and validity of prototype matching (Widiger 2011; Zimmerman 2011), a considerable body of research to support the reliability and validity of specific and explicit criterion sets (Zimmerman 2003), and no research that has compared directly the reliability or validity of independently administered prototype matching with specific and explicit criterion sets. Skodol (2010) cited in support of prototype matching studies conducted using the Personality Assessment Form (PAF). However, the authors of these studies in fact acknowledged that they used prototype matching only because at the time their study began semi-structured interviews to assess the DSM-III criterion sets were not yet available (Shea et al. 1987). Pilkonis et al. (2011) have since indicated their significant concern that prototype matching permits diagnosticians 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” (p. 73).

The only empirical support beyond the early PAF research was a validity study by Westen et al. (2006) and an interrater reliability study by Westen et al. (2010), both of which included fundamental methodological flaws. For example, in the case of the validity study, the clinicians who provided the prototype ratings also provided the criterion diagnoses, the latter even provided prior to their provision of the prototype ratings. Frankly, using this methodology, it would be difficult to obtain weak results, as the clinicians were simply confirming their own recently made judgments. If this criterion contamination was not problematic enough, the ratings were provided for patients the clinicians already knew extremely well (in treatment on average for 16 months), which is not the situation in which diagnostic criterion sets are typically used. With respect to the reliability study, these prototype ratings were obtained in the course of a 4.5 h standardized interview, inconsistent with the purported method of prototype matching. In addition, there was a clear possibility that the assessments were again not in fact blind to one another. The clinicians who provided the ratings were graduate students working together within a psychological clinic. It is not uncommon in such a setting for student clinicians to discuss amongst themselves their diagnostic impressions of new clients (and in some cases initial clients are discussed together at formal case meetings).

In response to the critiques of prototype matching (Widiger 2011; Zimmerman 2011), the PPDWG was compelled to abandon their proposal for prototype matching and to include instead diagnostic criterion sets (Siever 2011). In sum, it now appears that this major innovation for DSM-5 has been rejected. However, rather than work from the diagnostic criterion sets that were developed for DSM-IV-TR and have since been used in a substantial body of empirical research (e.g., Skodol et al. 2005), the PPDWG has apparently decided to construct brand new criterion sets by arbitrarily combining the self and interpersonal pathologies that they think will be specific to each respective personality disorder along with a list of traits they again think will likely be diagnostic of each personality disorder.

Deletion of Diagnoses

The PPDWG also intends to delete half of the diagnoses; more specifically, the dependent, narcissistic, paranoid, schizoid, and histrionic personality disorders. The primary reason for their deletion is to reduce diagnostic co-occurrence (Skodol 2010). Diagnostic co-occurrence has been a significant problem for the categorical diagnoses (Widiger and Trull 2007) but sacrificing fully half of them would seem to be a rather draconian approach for addressing this problem. In addition, it does not speak well for the credibility of the field of personality disorder to be so willing to sacrifice half of its coverage in order to address diagnostic co-occurrence, as if half of what we have been diagnosing and treating for the past 30 years was not worth the clinical attention (Widiger 2011). Persons will still have dependent, schizoid, paranoid, histrionic, and narcissistic personality traits despite their diagnoses being deleted (if not, then it is unclear why there is any need to include these traits within the dimensional model). Lack of adequate coverage has been a problem of comparable magnitude to diagnostic co-occurrence (Verheul and Widiger 2004). This problem will be magnified substantially in DSM-5.

In addition, significant questions have been raised with respect to the rationale for which diagnoses to delete. There does appear to be as much, if not more, empirical support for the narcissistic and dependent personality disorders (two diagnoses to be deleted) as there is for the avoidant and obsessive–compulsive personality disorders (Bornstein 2011; Ronningstam 2011). Zimmerman (in press) suggests that it is no accident that four of the five diagnoses being retained (i.e., avoidant, obsessive–compulsive, schizotypal, and borderline) were the focus of the CLPS project, spearheaded by the Chair of the PPDWG (Skodol et al. 2005).

Skodol et al. (2011) provides a review of the literature which they suggest indicates support for the decision to delete the dependent and narcissistic personality disorders in favor of the avoidant, obsessive–compulsive, antisocial, borderline, and schizotypal. However, even if one confines the decision to the studies cited by Skodol et al. one does not discover much support for the decision (Mullins-Sweatt et al. in press). For example, one of the reasons given for a weakness in the validity of dependent personality disorder was a difficulty in discerning its prevalence because the prevalence purportedly fluctuates widely from study to study. However, in the seminal review of epidemiology by Torgersen (2009), cited by Skodol et al. the fluctuation in prevalence was actually worse for the schizotypal (ranging from 0.0 to 3.2 across the studies that were considered), antisocial (0.0–4.5), borderline (0.0–3.2), avoidant (0.4–5.0), and obsessive–compulsive (0.0–9.3), the five to be retained, than it was for the dependent (0.4–1.8). Dependent personality disorder was also said to be associated with only moderate to low impairment in functioning, but its level of impairment has been consistently higher than has been obtained for the obsessive–compulsive in the studies considered by Skodol et al. (Mullins-Sweatt et al. in press). Finally, Skodol et al. indicated that dependent was one of the two least common personality disorders in the community, according to the review by Torgersen. However, this was not in fact the case. According to Torgersen’s review, with respect to the median rate across the studies he considered, dependent had a higher prevalence rate than schizotypal (and higher than three other personality disorders), and, when considering the pooled rate across these studies, a higher prevalence within the community than either schizotypal or borderline.

In any case, the decision of what to retain and what to delete might in fact be moot, as it now appears that there may not in fact be a personality disorders section, or at least if there is one, it could very well be reduced to a skeleton of its former self that is unlikely to survive any future harsh winter. Siever (2011) indicates that a representative of the PPDWG agreed with representatives of the schizophrenia disorders work group to move schizotypal personality disorder out of the personality disorders section into a new class of schizophrenia-spectrum disorders. Its primary coding will be as a schizophrenia-spectrum disorder, not as a personality disorder (the latter will only be noted parenthetically for historical purposes). A similar proposal is being pushed heavily for a shift of antisocial/psychopathic personality disorder into a new class of (child and adult) disruptive behavior disorders, wherein it would also receive its primary diagnostic coding and noted only parenthetically that it used to be classified as a personality disorder (Siever 2011). If these new proposals are enacted, the personality disorders section will be left with just three diagnoses (i.e., avoidant, obsessive–compulsive, and borderline), and it is difficult to imagine that the section could then survive (Widiger 2011).

Dimensional Trait Model

It is evident that the diagnosis and classification of personality disorder is shifting toward a dimensional trait model (Widiger and Simonsen 2005). It has in fact been suggested that the primary contribution of DSM-5 will be a shift of the entire diagnostic manual toward a dimensional model of classification (Regier 2008). This will be most clearly evident with the personality disorders, which will include a 6 (or 5) dimensional model of maladaptive personality, including 37 (or 25) lower-order traits that can be used to provide an independent description of each particular patient and/or be part of the diagnostic criterion sets for each respective personality disorder. Lanier et al. (2011) describe well many of the benefits and advantages of this shift.

Lanier et al. (2011) also document well how this proposal is well aligned with the five-factor model (FFM) of general personality structure. An integrative dimensional model of normal and abnormal personality offers quite a few benefits (Krueger and Eaton 2010; Widiger and Trull 2007). It addresses the many fundamental limitations of the categorical model (e.g., heterogeneity within diagnoses, inadequate coverage, lack of consistent diagnostic thresholds, and excessive diagnostic co-occurrence). It provides a more comprehensive and individually specific description of each patient’s normal and abnormal personality structure, thereby facilitating more precise and informative research concerning etiology and pathology, and more specific and distinct treatment decisions (Widiger and Mullins-Sweatt 2009). Finally, it transfers to the psychiatric nomenclature a wealth of knowledge concerning the origins, childhood antecedents, stability, and universality of the dispositions that underlie personality disorder (Widiger and Trull 2007).

Lanier et al. also indicate, however, that the authors of the DSM-5 dimensional trait model disavow some of the connection with the FFM; more specifically, that compulsivity is not a maladaptive variant of conscientiousness and oddity or peculiarity is not a maladaptive of openness (Clark and Krueger 2010; Krueger et al. 2011). The rationale for this position is unclear, as there is a considerable body of empirical research that supports the relationship of compulsivity to conscientiousness and oddity to openness (Widiger 2011). In addition, failing to acknowledge this continuum results in a model that lacks coherence or consistency, as if some dimensions of maladaptive personality are on a continuum with general personality structure (i.e., emotional instability or dysregulation, antagonism, detachment, and disinhibition) whereas others (i.e., compulsivity and oddity), are for no apparent reason qualitatively distinct from general personality structure. This has not been the position held previously by the proponents of dimensional models of personality disorder, including the authors of the model for DSM-5 (e.g., Clark 2007; Markon et al. 2005).

There are, however, important ways in which the dimensional trait model proposed for DSM-5 is different from the FFM. First, it does not actually include any normal personality traits, thereby failing to provide a truly integrative model. The DSM-5 proposal is confined to maladaptive personality traits, and thereby will not be able to identify the normal variants of the traits could in fact be quite useful, if not important, for treatment planning, such as openness that can suggest a responsivity to insight, reflective, and dynamic therapies, agreeableness and extraversion that can suggest a receptivity to group, marital, and other forms of interpersonal therapy, and conscientiousness that can suggest a willingness and ability to withstand the rigors of dialectical behavior therapy (Widiger and Mullins-Sweatt 2009). In addition, the model is entirely unipolar, failing to recognize the bipolarity of personality structure that has been empirically very well supported (Markon et al. 2005; Widiger 2011). The absence of this bipolarity contributes to the failure of the proposed model to recognize a number of important maladaptive personality traits, such as the glib charm and fearlessness of psychopathy (low neuroticism), gullibility and meekness of dependency (high agreeableness), and closedness to feelings of alexithymia (low openness).

In fact, like almost everything else for DSM-5, the dimensional model has apparently changed since the paper by Lanier et al. was accepted for publication. The model proposed at this current moment in time is a 5 domain model (emotional dysregulation, detachment, antagonism, disinhibition, and peculiarity), with 25 lower-order trait scales. The basis for this shift appears to be due simply to a recent factor analysis conducted by member(s) of the DSM-5 PPDWG (Siever 2011), rather than being guided by the considerable body of existing research. This may reflect the wider tendency of the DSM-5 process (Frances 2009) to allow work group members to rely on their own preferences and their own studies rather than seeking a more consistent historical continuity guided by a wider scientific literature.

Conclusions

In sum, Lanier et al. are indeed correct that DSM-5 personality disorders are likely to be much different than the DSM-IV-TR personality disorders. Some of these significant changes could reflect major improvements in how disorders of personality are conceptualized and diagnosed. However, the proposals vary considerably in the extent to which they have compelling empirical support. Even a member of the DSM-5 PPDWG has opined that “the DSM-5 proposal is a disappointing and confusing mixture of innovation and preservation of the status quo that is inconsistent, lacks coherence, is impractical, and, in places, is incompatible with empirical facts” (Livesley 2010, p. 304), characterizing the overall effort as an expression of “incoherence and confusion” (p. 304). The major accomplishment of the fourth edition of the APA’s diagnostic manual was not in the development of surprising new content but rather in the careful, cautious, and systematic method with which it was constructed. The authors of the forthcoming fifth edition may have turned this priority on its head, emphasizing instead radical changes without first conducting careful, systematic, thorough, or objective reviews of the scientific literature.