The target article by Boorsboom et al. infuses new ideas into a longstanding debate on the definition, classification, and explanation of mental disorders. It has the potential to create novel ways of mapping symptom structures, understanding comorbidity, and tracking the emergence and development of symptoms. We argue that, while the symptom network model (SNWM) can play an important role in identifying patterns of symptoms, it offers less when it comes to their explanation.

In their article, Borsboom et al. present five, overlapping, theoretical and empirical arguments critiquing the neurocentric, reductionist approach to defining and explaining mental disorders, and propose that a SNWM approach should be adopted in its place. In essence, they argue that:

1. 1. Mental disorders are not disease entities, but are best understood as dynamic networks of symptoms that trigger each other rather than as “things.”

2. 2. Contrary to the claim of neurobiological conceptions of mental disorders, the mind cannot be reduced to the brain (e.g., “mental disorders are not brain disorders at all,” sect. 1, para. 5).

3. 3. Understanding the intentionality of symptoms (what symptoms mean at a phenomenal level) is key.

4. 4. External-contextual factors are important causes of mental disorders and initially activate symptom networks that then become self-sustaining.

5. 5. In light of the above arguments, multilevel explanatory strategies that cover neurobiological, phenomenological, and social-cultural domains are required in order to describe and explain symptom networks.

Borsboom et al. conclude that SNWMs can deal with these challenges and define, classify, and explain mental disorders more adequately than reductionist neurobiological theories.

A first critical point is that these arguments are only loosely aligned, and although collectively seeming to provide a powerful challenge to neurological conceptions of mental disorders, do not in fact do so. The reason for this is that they individually fail to provide good reasons for adopting a symptom network approach to explaining psychopathology. In brief:

1. (a) The primary causes of mental disorders could be biological in nature without being disease entities.

2. (b) Reductionism is a metaphysical argument, not an epistemological (explanatory) one. Therefore, it is not seriously considered by contemporary cognitive neuroscientists, who agree that personal and subpersonal psychological levels of explanation provide unique and valuable insights into the mind (Eliasmith Reference Eliasmith2013).

3. (c) Privileging the intentionality of symptoms on a priori grounds is to commit oneself to folk psychological conceptions of the mind and risk begging important theoretical questions.

4. (d) Multilevel explanations can be developed within the context of a view of the mind as a material system. In other words, a strategy of explanatory pluralism (for pragmatic reasons) can be neutral with respect to the nature of the mind from a metaphysical perspective.

In our view, the failure to be clear about the limits of these distinct arguments means that there is a temptation to structure the debate between neurobiological and intentional level explanations as one between reductionist versus non-reductionist approaches. This is not helpful and represents a false dichotomy.

A second critical point is that the SNWM approach is best viewed as simply one phase of scientific inquiry, and requires supplementing by other types of methodological and theoretical models. There are a number of different aims evident in scientific practice, including prediction, description, classification, and explanation – both causal and compositional (Craver & Kaplan, Reference Craver and Kaplanin press). Different kinds of models are constructed in service of these aims. Phenomenal models are essentially descriptive in nature and aim to capture patterns such as co-occurring events or mental states and behavior (Hochstein Reference Hochstein2016). Explanatory models set out to depict the components, their relationships, and organisation in mechanisms that underlie phenomena (Craver & Kaplan, Reference Craver and Kaplanin press). For example, a phenomenal model of depression seeks to describe the symptoms associated with the syndrome and their temporal relationship to one another, while explanatory, mechanistic models provide insight into their underlying constituents, processes, and organization (e.g., serotonin dysregulation, neural activation patterns, attentional biases, negative self-evaluations; Beck & Bredemeier Reference Beck and Bredemeier2016).

In our view, the symptom network model is usefully construed as a phenomenal model and therefore is best suited to detecting patterns among symptoms, as opposed to representing the mechanisms that constitute them (Craver & Kaplan, Reference Craver and Kaplanin press; Hochstein Reference Hochstein2016). Identifying the covariance between symptoms and the order in which they appear is an important scientific task and provides a description of phenomena that can subsequently be a focus of explanation. Symptom network models have a crucial role to play in the scientific inquiry process, but a limited one.

Thinking of the SNWM in these terms allows room for developing constitutive and etiological explanations of each symptom (nodes in the network) once the relevant patterns have been identified. Symptoms are complex structures constituted by entities and processes at different levels of organization. Thus, it is reasonable to construct explanatory models, each directed at different aspects of symptoms, and collectively providing an overall explanation of them (i.e., phenomena). These mechanistic models shed light on the way different symptoms exert their effects and influence each other.

A final critical point is that Boorsboom et al. overlook the plurality of models required to engage in research. Even from a SNWM perspective, a family of network models is needed to identify and model relationships among symptoms, each with a different goal: modeling symptom structure within mental disorders, symptoms clusters (irrespective of diagnosis), temporal shifts (e.g., progression within individuals), and so on. Networks of explanatory and descriptive models are necessary to provide a comprehensive explanation of symptoms and their grouping into disorders, each focusing on different types of factors and processes.