Historically, biomedical research has oscillated between two principle scientific paradigms: mapping of phenomena to remote causes versus rejecting the intelligibility of such remote causes and pursuing a taxonomy on the level of the phenomenon (Zachar & Kendler Reference Zachar and Kendler2017). Most medical fields have settled successfully on the former paradigm, but in psychiatry, the pendulum continues to swing. The rejection of a biological reduction of psychiatric phenomena and the retreat to discretional mapping of symptoms, as follows from Borsboom et al.’s argumentation in the target article, is in line with the currently observed swing-back of the pendulum to the latter paradigm. This motion is motivated by the valid negative critique of the limited therapeutic success that biological research in psychiatry has had and is conceptually grounded in non-reductive materialism.

Specifically, the authors argue: “Currently, there is no compelling evidence for the viability of reducing mental disorders to unique biological abnormalities” (target article, sect. 1, para. 4). They conclude that this reasons against distinct biological causes for mental disorders. Although we agree with the premise, both from an empirical and a theoretical perspective, we do not agree that the conclusion is justified. Both historical evidence from biological reduction in other fields of medicine (Kotchen Reference Kotchen2011) and the principle of minimal assumptions (Ockham's razor) favor the alternate conclusion that it is not the assumption of biological causes, but the artificial and arbitrary labeling of mental disorders that is problematic. Artificial, ever-changing definitions are unlikely to lead to a biologically coherent cause – or biological correlate (Katahira & Yamashita Reference Katahira and Yamashita2017). We agree with Borsboom et al. that a research paradigm that tries to unravel the causes of such artificial entities is flawed by essentialism; however, shifting the focus from disorders (i.e., symptom-aggregates) to symptoms inherits the problem of essentialism: For what is a symptom? A symptom is merely a verbal subjective or behavioral consequence of a (dysfunction in a) given process/system. In analogy to other organ systems, the relationship between disturbance in such a process and reported symptoms may be nonlinear and heterogeneous/pleiotropic. In accordance with the authors, this relationship might set off a cascade of interactions that, together, affect the course of illness. In this sense only, the symptom-networks approach, as advocated by the authors, precludes biological reduction (as does the partial-reduction approach that the authors propose). Instead of abandoning biological reduction altogether, or continuing the hunt for biological substrates of artificial entities, why not follow the formula that has proven most successful in all other medical disciplines?

For example, the cardinal symptom in cardiology, chest pain, does not reflect any relevant biological substrate, but emerges from a variety of causes (from heart ischemia to intercostal neuralgia) (Lenfant Reference Lenfant2010). What has paved the road to reduction in this discipline was a paradigm shift toward the concept of processes (physiology and pathophysiology) (Granger et al. Reference Granger1998). Symptoms, formerly the sole means of establishing a diagnosis, now merely serve as a guiding torch, but the characterization of patients takes place directly at the level of these processes (e.g., ECG). An approach based on the understanding of how a quantitatively measurable physiological process (myocardial oxygen delivery and extraction) can be disturbed (sclerotic plaque) and how this unreliably maps to specific symptoms (from none to severe chest pain or abdominal symptoms), has only recently started to be developed in psychiatric research (Friston et al. Reference Friston, Redish and Gordon2017; Peters et al. Reference Peters, McEwen and Friston2017; Stephan et al. Reference Stephan, Manjaly, Mathys, Weber, Paliwal, Gard, Tittgemeyer, Fleming, Haker, Seth and Petzschner2016).

Translated into the context of psychiatry, these processes will need to reflect relevant classes of computations that the brain performs to maintain the organism's homeostasis (Friston Reference Friston2010). The definitions of such processes (e.g., reward prediction processing, salience monitoring, or executive functioning), as continuous functional entities cannot be derived directly from symptom properties. Instead, they need to be informed by comparative physiological and functional anatomical studies across species and through investigations in healthy individuals.

One model of such a process that has attracted recent attention is the process of reward prediction, representing a proposed fundamental neural computation that takes place at multiple hierarchical levels (from sensory input to higher-order associations) and serves to update the brain's model of the world in the face of new evidence (sensory input) (Keiflin & Janak Reference Keiflin and Janak2015). Although these processes span a space in which subjective, reportable experience might correlate with individual variance along these axes (e.g., prediction error and feeling of surprise), the success of reduction will have to be evaluated by the utility of direct assessment and manipulation with regard to relevant outcomes.

In Figure 1, we illustrate exemplary dimensions of a putative process and how we envision its non-unique relationship to commonly used symptoms.

Figure 1. The putative process of reward prediction, with three exemplary dimensions/axes (out of many). The x axis depicts the accuracy of a prediction (i.e., the tendency to underestimate or overestimate an outcome). The y axis reflects the precision of a prediction (i.e., ranging from a point prediction to broad interval predictions), and the z axis represents the strength of the error signal that drives learning (i.e., from weak to strong) An overestimation of positive outcomes combined with a high precision and a weak error signal (see black dots for exemplary individual scores) may lead to dysfunctional behavior such as gambling. Importantly, the resulting behavior and subjective interpretation (e.g., symptoms) can range across a broad spectrum and might not be unique to this process (see insert at right upper corner).

Whether a process-centered approach will yield a complete description of mental states (in terms of content) in material terms within an individual's brain is not relevant for psychiatric research or praxis. Relevant are the dimensions of mental states, such as anxiousness, vigilance, or relation to reality. Importantly, acute manipulation along these dimensions – for instance, pharmacologically – is possible (e.g., benzodiazepines, propofol, or LSD, respectively).

Conceptualizing psychiatric phenomena as the result of disturbances in functional brain processes allows us to overcome another limitation that Borsboom et al. suggest as intangible with biological reduction: accounting for social and normative factors. The authors exemplify this with gambling, which is strongly contingent on environmental (presence of gambling booths) and normative factors. When we regard gambling not as the substrate of reduction but as a phenomenon that can emerge from dispositions in different processes (e.g., reward processing, executive functioning), such an account becomes tangible. One can now investigate how environmental and societal normative factors shape these processes through an interaction between genomic setpoints and outside influences (e.g., stressors) on cellular, circuit, and behavioral levels. Such an integrated research paradigm was put into practice recently in the National Institute of Mental Health Research Domain Criteria (Insel et al. Reference Insel, Cuthbert, Garvey, Heinssen, Pine, Quinn, Sanislow and Wang2010) and the ROAMER initiative (Schumann et al. Reference Schumann, Binder, Holte, de Kloet, Oedegaard, Robbins, Walker-Tilley, Bitter, Brown, Buitelaar, Ciccocioppo, Cools, Escera, Fleischhacker, Flor, Frith, Heinz, Johnsen, Kirschbaum, Klingberg, Lesch, Lewis, Maier, Mann, Martinot, Meyer-Lindenberg, Müller, Müller, Nutt, Persico, Perugi, Pessiglione, Preuss, Roiser, Rossini, Rybakowski, Sandi, Stephan, Undurraga, Vieta, van der Wee, Wykes, Haro and Wittchen2014). In conclusion, the pendulum continues to swing, and there is not yet reason to abandon the effort of biological reduction that has been most fruitful in all other fields of medicine.