In the past, one might have argued that the differential diagnosis of specific personality disorders made little difference, that it wasn't a useful clinical guide for individual patients. However, research has demonstrated differences in clinical course and prognosis among the several personality disorders, and the separate categories have been useful to the growing body of research on therapeutics.

As we move toward DSM-5, it is clear that the clinical and research communities view personality disorders differently. The clinical community wants a system that is practical and workable in the real world and that focuses on the essence of each category. The research community wants to capture as much information as possible and to emphasize precise boundaries of categories rather than reifying core syndromes that may have more to do with tradition or theory than with patients.

In this issue of the Journal, Westen et al. (1) enter the fray with the goal of “bridging” science and practice. They claim that they are developing a “taxonomy” (the term Linnaeus introduced for classifying living things according to their natural relationships). DSM is more modest, claiming only to be a “nosology” (a classification of diseases). In fact, “nomenclature” (a system of names) might be even more appropriate. Westen and colleagues' important study is the most recent in a 15-year program of research that has established their position as an exemplar representing one important position in the dialogue of personality disorder diagnosis.

Westen et al. argue that personality disorders are primarily clinical concepts. The individual disorders are syndromes—clusters of meaningfully related characteristics that are recognized as syndromic entities, not as collections of independent phenomena. In explaining the concept, the authors use the metaphor of face recognition; it is relatively easy when we see a whole face but much more difficult if we are presented with an assortment of eyebrows, noses, chins, eyes, and mouths.

Westen et al. have developed prototypic descriptions of eight personality disorders, two of “neurotic styles,” and one of personality health. In order to do this, they did not study patients directly but rather studied the way that clinicians conceptualize their patients.

The authors developed a random sample of psychiatrists and psychologists and asked each to describe a patient using items selected from a set provided by the authors. They applied a statistical strategy that identified clusters of patient descriptions and then transformed the most characteristic descriptors of each cluster into a brief narrative. The article in this issue expands on earlier work by Westen et al. with a larger pool of clinicians and patients (N=1,201), fewer exclusive criteria (no longer requiring a DSM personality diagnosis), and further refinement of the list of items.

Westen et al. emphasize that their system is “empirically derived,” unlike the theory-driven tradition-based systems of their predecessors. This is true, up to a point (perhaps as true as possible). There are three inputs to their data: the patients, the clinicians, and the descriptive items. The patients were selected because the treating clinicians view them as having personality problems (generating a “sicker” sample than I would have predicted; one-third of the patients had a previous hospitalization). The clinicians, not the patients themselves, are the real research subjects, and like all clinicians, they are contaminated by the concepts and theories that they bring with them. However, the authors went to great lengths to neutralize this factor by using a wide range of clinicians from a variety of theoretical and professional backgrounds; less than 25% self-identified as “psychodynamic”—the investigators' orientation. The items were assembled over time, largely refined by their psychometric properties and by the consumer responses of the clinicians. The current iteration of the instrument has 200 items, of which 90% have been only minimally revised from the last version. Accepting that the authors are using verbal reports of observations of psychological functioning—one human describing another—this would seem as objective and empirical as one is likely to get. However, inevitably, it sees patients through two lenses—the clinicians' and those of the researcher who selected the items.

Most of the criticism directed at Westen and colleagues' research has not been that they are doing what they do poorly but rather that they are doing it at all. One group, composed largely of research psychologists whose work has been with healthy subjects rather than patients, argues that personality comprises multiple traits that are each distributed continuously and thus are best captured by a multidimensional system. In this group, categories are an unfortunate attempt to impose biomedical concepts on a psychological domain in which they do not fit. Westen et al. reply that trait psychology was developed and is most useful in the study of normal personality. Although trait psychology can be used to describe personality disorders, the result is awkward and clinically clumsy. A second critique comes largely from clinical researchers who value the precision of explicit criteria and diagnostic algorithms and fear that a shift to prototypes would emphasize the core rather than the rules for decisions at the boundaries. This clinical research community is also understandably concerned that any change in the diagnostic system will threaten the value of years of research conducted under previous systems.

How do clinicians respond to their system? Here I was able to personally contribute some empirical data, albeit with an N of only one. As a clinician I was interested in the diagnostic prototypes that emerged from the study. Most seemed familiar and unsurprising, although the system was easier to use than lists of criteria and far more clinically friendly than multiple scales of traits. I read the descriptions carefully to see what might be confusing or contradictory to my personal clinical wisdom. I found only three, all relatively minor.

1. The authors used a factor-analytic approach to sort the eight disorders that they identified into three superordinate “spectra” (replacing the “clusters” of axis II in DSM-IV) and then provided brief descriptions of each of the spectra. They consider a schizoid-schizotypal category (merging these categories, as clinicians are prone to do and researchers are not) as part of the “internalizing spectrum.” However, they state that individuals in this spectrum “experience chronic painful emotions, especially depression and anxiety,” while schizoid-schizotypal individuals are described as having “a limited or constricted range of emotions.” This seems inconsistent—“chronic painful” or “limited or constricted.” Looking further, I noted that in their study of diagnostic intercorrelations, schizoid-schizotypal diagnoses had much lower correlations with other diagnoses in their own spectrum than was true for any other personality disorder. In brief, it doesn't quite fit into its spectrum.

2. The description of the hysteric-histrionic personality includes the phrase “tend to be suggestible or easily influenced by others.” This is a traditional view, but I believe it is incorrect. These patients claim to be suggestible in order to deny motivation and agency, but closer study reveals that they are skillful in selecting who will make suggestions, eliciting their desired suggestions, and only then appearing to “follow” them. This may illustrate a limitation of this research strategy—it studies clinicians' perceptions of patients, not the patients themselves, and if clinicians have a persistent bias that stems from misleading notions that they have been taught, and if the same notions influenced the investigators who selected these 200 items, there is no way for this approach to correct the error or even to identify it.

3. The authors developed an 11th prototype labeled “personality health.” For the most part it read like a scout manual, as expected. One item (“outgoing”) seems more like a cultural value, perhaps a common one among mental health professionals, rather than a component of health. I know inward-directed individuals who seem quite healthy and outgoing people who do not.

The gap between researcher and practitioner in personality disorders may be fundamental—the diagnoses are used for different purposes. Westen et al. have provided a state-of-the-art strategy for constructing categories that reflect how clinicians think and that clinicians will find friendly to use. The architects of DSM-5 will have to decide how it should resolve the tensions between the clinical and research communities and their different goals in using the nosology.