First, we question whether Cramer et al.'s proposed network model can provide a viable scientific foundation for investigating comorbidity without invoking latent variables in some form. Second, the authors' claim that the network perspective is radically different from a latent variable perspective rests upon an undemonstrated premise. Without being demonstrated, we think the premise is potentially misleading.

Cramer et al. contrast two possible explanations for psychological symptoms: latent variables (i.e., specific cause) versus a network of causality between symptoms. There is a third explanation: The reason for comorbidity and the reported network structure of psychological symptoms is that the underlying biological cause is a psychoneuroimmunoendocrine information network which, when dysregulated, leads to several maladaptive psychological and somatic symptoms.

Latent variable modeling has revealed important conundrums in the DSM classification system. We agree that the network perspective has potential to inspire new insights and resolve some of these conundrums. We note, however, that alone it cannot really help us understand etiology. Etiology, not comorbidity, is the fundamental question.

Cramer et al. articulate a novel perspective on comorbidity. However, their network models must be compared with more parsimonious latent variable models before conclusions can be drawn about network models as plausible accounts of comorbidity. Latent variable models have proven generative in studying psychopathology and its external correlates, and we doubt network models will prove as useful for psychopathology research.

Cramer et al. present a thoughtful application of network analysis to symptoms, but certain questions remain open. These questions involve the intended causal interpretation, the critique of latent variables, individual variation in causal networks, Borsboom's idea of networks as measurement models, and how well the data support the stability of the network results.

In the target article, Cramer et al. suggest that diagnostic classification is improved by modeling the relationship between manifest variables (i.e., symptoms) rather than modeling unobservable latent variables (i.e., diagnostic categories such as Generalized Anxiety Disorder). This commentary discusses whether symptoms represent manifest or latent variables and the implications of this distinction for diagnosis and treatment.

Cramer et al. present an original and interesting network perspective on comorbidity and contrast this perspective with a more traditional interpretation of comorbidity in terms of latent variable theory. My commentary focuses on the relationship between the two perspectives; that is, it aims to qualify the presumed contrast between interpretations in terms of networks and latent variables.

Cramer et al. persuasively conceptualize major depressive disorder (MDD) and generalized anxiety disorder (GAD) as network disorders, rejecting latent variable accounts. But how does their radical picture generalize across the suite of mental and personality disorders? Addictions are Axis I disorders that may be better characterized by latent variables. Their comorbidity relationships could be captured by inserting them as nodes in a super-network of Axis I conditions.

In developmental psychopathology, differentiating between the coexistence and the clinical entity of two problem areas is of utmost importance. So far, logistic regression analysis has already provided helpful answers, as shown in studies on comorbidity of tic disorders. While the concept of bridging symptoms may be investigated adequately by both logistic regression and the network approach, the former (latent variable) seems to be of advantage with regard to the problems of multiple comorbidities and development.

We agree with Cramer et al. that pure cases of behavioral disorders with no symptom overlaps are rare. However, we argue that disorders do exist and the network idea is limited and limiting. Networks of symptoms are observed mainly at behavioral levels. The core deficit is commonly at the cognitive or brain levels, and there the story is completely different.

As psychiatric illnesses have correlates in the brain, it is surprising that Cramer et al. make almost no reference to the brain's network character when proposing a network approach to comorbidity of psychiatric diseases. We illustrate how data from combined neuropsychological and functional and structural brain-imaging investigations could inform theoretical models about the role played by overlapping symptoms in the etiology of psychiatric comorbidity and the pathways from one disorder to another.

We suggest that the network approach to comorbidity (Cramer et al.) is best examined by using longitudinal, multi-measurement, intra-individual data. Employment of time-series analysis to the examination of the generalized anxiety disorder and major depressive disorder comorbidity enables a detailed appreciation of fluctuations and causal trajectories in terms of both symptoms and cognitive vulnerability.

Cramer et al's proposal to view mental disorders as the outcome of network dynamics among symptoms obviates the need to invoke latent traits to explain co-occurrence of symptoms and syndromes. This commentary considers the consequences of such a network view for genetic association studies.

Cramer et al.'s article is an example of the fruitful application of complex dynamic systems theory. We extend their approach with examples from our own work on development and developmental psychopathology and address three issues: (1) the level of aggregation of the network, (2) the required research methodology, and (3) the clinical and educational application of dynamic network thinking.

We welcome network theory as a tool for modelling the multi-directional interactions that characterise disease. However, we feel that Cramer et al. have neglected one important aspect: how diseases change over developmental time. We discuss principles such as fan in, fan out, bottlenecks, and common pathways, and argue that modelling these developmental aspects can be vital, particularly in deriving properly targeted treatments.

Cramer et al.'s network approach reconceptualizes mental comorbidity on the basis of symptom space originating from psychometric signatures. We argue that the advantages of this approach need to be regarded in the context of the multi-level functional organization of the neural substrate, ranging from neurogenetic to psychometric. Neuroelectric oscillations are proposed as a level-integrating principle.

Cramer et al.'s critique of latent variables implicitly advocates a type of scientific anti-realism which can be extended to many dispositional constructs in scientific psychology. However, generalizing Cramer et al.'s network model in this way raises concerns about its applicability to psychopathology. The model could be improved by articulating why a given cluster of symptoms should be considered disordered.

The majority of commentators agree on one thing: Our network approach might be the prime candidate for offering a new perspective on the origins of mental disorders. In our response, we elaborate on refinements (e.g., cognitive and genetic levels) and extensions (e.g., to Axis II disorders) of the network model, as well as discuss ways to test its validity.

Although cognitive scientists have learned a lot about concepts, their findings have yet to be organized in a coherent theoretical framework. In addition, after twenty years of controversy, there is little sign that philosophers and psychologists are converging toward an agreement about the very nature of concepts. Doing without Concepts (Machery 2009) attempts to remedy this state of affairs. In this article, I review the main points and arguments developed at greater length in Doing without Concepts.

I raise two issues for Machery's discussion and interpretation of the theory-theory. First, I raise an objection against Machery's claim that theory-theorists take theories to be default bodies of knowledge. Second, I argue that theory-theorists' experimental results do not support Machery's contention that default bodies of knowledge include theories used in their own proprietary kind of categorization process.