In this issue, Kendler et al. (1) describe genetic and environmental factors that underlie the significant relatedness observed among multiple axis I and axis II mental disorders. Specifically, they report that four potent genetic factors, as well as three environmental factors, each contribute to clusters of psychopathology listed in DSM-IV. They describe the genetic factors as axis I internalizing, axis II internalizing, axis I externalizing, and axis II externalizing. It should be noted, however, that schizophrenia and other psychotic disorders were excluded from this study.

Research reports of this kind can seem quite arcane and far removed from clinical practice. In fact, the type of analysis on which Kendler and colleagues build may exert a substantial influence on the currently ongoing revisions of DSM-IV and ICD-10. Historically, the most compelling reason to search for shared risk factors underlying mental disorders is the extraordinarily high frequency of comorbidity among DSM disorders (2). The observed rate of co-occurrence among some disorders is so high that it demands explanation. Hypotheses, which are not mutually exclusive, include the possibilities that 1) the co-occurring disorders represent divergent expressions of at least partly shared risk factors, 2) one disorder plays a causal role in another, just as nicotine addiction plays a causal role in lung cancer, and 3) disorder boundaries in the DSM system are badly enough drawn that at least some of the comorbidity is an artifact of giving one illness multiple names.

Kendler and colleagues made important early contributions to the analysis of comorbidity by looking at disorders pair-wise. For example, they found (3) that generalized anxiety disorder and major depression have significantly overlapping genetic and environmental risk factors, likely explaining why so many patients in the clinic have both disorders. Krueger (4), Krueger and Markon (5), and other research groups made signifi-cant advances by performing multivariate analyses that yielded models of the relationships among multiple mental disorders that co-occurred at frequencies exceeding what would be expected from their population prevalences. Out of these analyses emerged hypothesized liability factors that explain the relationships not only among pairs of disorders, but also among clusters. The most widely replicated clusters have been given the names "internalizing disorders" (mood and anxiety disorders) and "externalizing disorders" (conduct disorder, oppositional defiant disorder, antisocial personality disorder, substance use disorders, and in many studies, attention deficit hyperactivity disorder). In short, these analyses predict that anxiety disorders are likely to co-occur with major depression or dysthymia, but not with antisocial personality disorder. The "internalizing" and "externalizing" terminology has accreted diverse theoretical meanings over the years, raising questions as to whether these clusters deserve new names, but the basic observation that specific disorders co-vary within these clusters appears to be quite robust.

The emergence of clusters from multivariate analyses immediately raises the question of whether there are underlying common risk factors. That question has given rise to a large literature examining genetic and environmental risk factors that might explain why certain disorders co-occur at high rates while others do not. That literature cannot be reviewed in a short commentary, but a salient example is a study published recently in the American Journal of Psychiatry (6) that found a highly heritable general liability factor that accounted for the transmission of externalizing psychopathology across generations. What Kendler et al. have accomplished in the study reported in this issue is to analyze genetic and nongenetic risk factors across the internalizing and externalizing clusters at once and, at the same time, to study (albeit with some methodological complexities) the 10 DSM-IV personality disorders.

Kendler et al. based their analyses on interviews with 669 monozygotic and 377 same-sex dizygotic twin pairs. With the standard assumption that same-sex twin pairs raised together share their early environment, a higher concordance rate for a trait among monozygotic twin pairs (who share 100% of their DNA sequences) than among dizygotic twin pairs (who on average share 50%) indicates a significant role for genes in the expression of the trait, in this case a mental disorder. Of course, twin studies do not yield the individual DNA sequence variations that contribute to risk; molecular studies that attempt to do so are still in relatively early stages for mental disorders and are revealing a remarkable degree of genetic heterogeneity and complexity. The advantage of twin studies is that they are agnostic about the actual genes involved but permit the computation of aggregate genetic risk, shared environmental risk, and unshared environmental risk plus chance. As noted already, Kendler et al. identified four genetic risk factors, each underlying a cluster of psychopathology, of which axis I and axis II internalizing disorders were moderately correlated with each other, as were axis I and axis II externalizing disorders.

I will not delve deeply into the conclusions drawn with respect to the axis II (personality) disorders, but I think that it is important to highlight several points. As in many other studies, including the recent one by Bornovalova et al. (6), antisocial personality disorder clusters with the axis I externalizing disorders and not with the personality disorders. Second, it is difficult to interpret findings with respect to schizotypal personality disorder given the lack of inclusion of schizophrenia-related diagnoses in the twin samples. The diagnosis of schizotypal personality disorder initially emerged from the examination of the nonpsychotic but symptomatic members of families with schizophrenic probands. Indeed, Kendler and colleagues have published compelling research elsewhere (7) that might argue for an independent cluster of psychopathology based on schizotypy. How this squares with the current study is not at all clear. The most difficult problem for an approach that starts with DSM disorders as building blocks is precisely its reliance on unvalidated DSM-IV disorder definitions. This concern is most pointed for some of the personality disorders; it is unclear how DSM-5 will ultimately treat the personality disorders, but at least four, and as many as five, of the existing disorders may disappear from the manual, in part because of a lack of coherence and a lack of validation even with respect to other diagnoses.

Modern classifications of mental disorders containing operationalized diagnostic criteria have had an important beneficial effect on interrater reliability. From the beginnings of modern approaches to psychiatric diagnosis (8—10), however, many investigators have attempted, largely without success, to establish validity for the existing disorders as "natural kinds" or to refine or subdivide them in a search for homogeneity. Over time, instead of validation, severe problems with the existing classifications have become increasingly obvious. These include not only the co-morbidity problem already discussed, but also the need for clinicians to rely with uncomfortable frequency on "not otherwise specified," or NOS, diagnoses (11, 12). The need to rely on NOS diagnoses suggests that the DSM system has a fundamental problem of overspecification: patients fully meeting criteria for DSM-IV disorders, especially for a single DSM-IV disorder, may well be the minority of individuals seen in clinical settings. To compound these problems so evident to clinicians, scientific studies, most notably genetic studies, have found shared genetic risk factors across putatively distinct disorders. These shared risk factors have been identified both by twin studies that identify aggregate genetic risk (3) and, increasingly, by molecular studies, which, for example, have found shared genetic contributions to schizophrenia and bipolar disorder (13).

From the foregoing brief description, it should be no surprise that the individuals tasked with producing DSM-5 and ICD-11 are bedeviled by the limited scientific knowledge in our possession to permit better definitions of individual disorders (14, 15). The article by Kendler et al. in this issue may give additional momentum to a strategy that I confess to having advocated (15), which is to put a significant new focus on the large groupings of disorders in DSM-IV and ICD-10 and to encourage investigators to work across the boundaries of individual disorders with the goal of reanalyzing, from the bottom up, where disorder boundaries should be drawn. It may well turn out that major sections of DSM-5 will correspond to the internalizing and externalizing disorders found in the work of Kendler et al. (1) and other groups.

For a clinician, the work by Kendler and colleagues also has meaning in practice with individual patients. For many patients, the quest for a definitive diagnosis is daunting, as different psychiatrists and psychologists posit various personality disorders, mood disorders, and anxiety disorders as the "cause," or at least the label, for their difficulties in life. For the patient who inquires about the rationale behind a series of seemingly different diagnoses, the article by Kendler et al. offers some explanation. Underlying common genetic and environmental factors appear to explain the overlap between members of major groups of illnesses. Individual differences, not shared in the family upbringing, reflect differences in circumstance that may change over time and are responsible for the expression of the illness at any particular stage of life. Thus, as an individual matures and experiences life for him- or herself, the manifestation of any behavioral trait, including mental illness, reflects the unfolding of that person's individuality and unique life story.