Secondary Logo

Journal Logo

Focusing on the Biological System Instead of the Biological Level in Psychiatry Research

Öngür, Dost MD, PhD

doi: 10.1097/HRP.0000000000000238
Disruptive Innovation

From Harvard Medical School and McLean Hospital, Belmont, MA.

Correspondence: Dost Öngür, MD, PhD, McLean Hospital, 115 Mill St., Belmont, MA 02478. Email:

Early in my psychiatry training, I had experiences that I later discovered were all too common among psychiatrists. Many of the patients that I saw met criteria for multiple psychiatric disorders, but their psychopathology appeared continuous, not forming distinct disorders. Almost all of my patients experienced a mix of anxiety, mood, psychotic, and addictive symptoms at some point during their illnesses, albeit to varying degrees. I also noticed that patients with classic presentations of syndromes were a minority and that most of my patients’ presentations evolved over time. Finally, it was common for me to see multiple patients who met criteria for the same syndrome but whose symptoms, clinical course, and treatment response were dramatically different from one another. It was not possible for me to identify the “essence” of any given syndrome that could aid reliable recognition.

We face an epistemological problem in psychiatry. Our existing psychiatric diagnoses do not address comorbidity, heterogeneity, and variability. Although complexity is commonplace in all of medicine, knowledge of pathophysiology in many other specialties allows doctors to parse clinical heterogeneity, predict treatment response, and advance translational research using markers and measurements. This is not the situation in psychiatry. In our field there is no easy way to pit one conceptual framework against another in terms of how well they explain the clinical reality. Instead, we use consensus building among experts and iterative collection of clinical data to try to make progress. As a result, the nosological system in psychiatry has come under heavy fire,1 and its shortcomings hamper our ability to make progress in research as well in as clinical care.

In fact, most agree that biological psychiatry research has not made meaningful progress sufficient to affect clinical care in several decades. Most biological studies identify subtle brain abnormalities of unclear clinical relevance. Non-replication is the rule, not the exception.2 This is striking because the phenotypes we study are by no means subtle: if we asked lay people to meet a stranger for the first time and tell us whether they think the individual is psychiatrically healthy or has schizophrenia, we would achieve excellent sensitivity and specificity within a few minutes. But the most sophisticated biological measures have struggled to satisfy much more lenient criteria.

So, how can we make better progress toward understanding psychiatric disorders? I suggest the answer lies in the nature of the brain as a dynamic, complex system. Its functioning is characterized by ongoing spontaneous and oscillatory activity that is also constantly modulated by sensory inputs and motor outputs that are fed back into the system.3 Static brain features can rarely produce behavior; rather, all brain functions are the result of modulation in patterns of brain activity. Static measures such as, say, cortical thickness or GABA concentrations, or measures averaged over time, such as resting-state connectivity metrics, can at best be downstream results of abnormal dynamic patterns and not the problems in and of themselves. Human experience does not take shape in a single level of description but rather in dynamic system properties. We need to envision ways to study these properties in ways that encompass all levels of biology (Figure 1).

Figure 1

Figure 1

Like all dynamic, complex systems, the brain maintains critical functional parameters within homeostatic ranges. This generates a predictable internal milieu built to receive and process expected inputs and to generate a repertoire of appropriate actions in ongoing fashion—for example, through predictive coding of changes in the environment.4 The brain contains countless feedback loops and compensatory mechanisms at all levels and timescales to maintain this milieu. When homeostasis is disrupted, it must be due to effects that the system did not fully anticipate and that the organism must absorb.5 In fact, the importance of a parameter can be gauged from the brain’s ability to maintain it within a narrow range. In a framework of hierarchical control of the system, less important parameters can be “sacrificed” to move outside of homeostatic range in response to perturbation so that other, more important, parameters remain within range.6 If the disruption is severe enough, the system can lose its ability to function altogether. In the brain, information processing can become unstable and chaotic (or alternatively, fixed and static) when important parameters remain outside their usual dynamic ranges. These conditions are reminiscent of “failure modes” described by engineers in complex systems.7

Disorders of cognition and emotion are manifestations of systemic abnormalities. It is reasonable to assume that these abnormalities must have underlying molecular, cellular, or circuit correlates. But symptom formation in psychiatric disorders can arise only in spontaneous brain activity and its context-dependent modulation; that is, molecular lesions are not sufficient if they do not affect brain activity. The brain is expert in absorbing disruptive influences and repositioning itself to function optimally. If neurochemical levels or regional brain activity are functioning outside of homeostatic ranges and their dynamic modulation has been disrupted, there must be compelling reasons. Adverse influences unfolding over time and interacting with individual predispositions make it harder and harder for the brain to accommodate, and ultimately lead to abnormal functioning. Others have used this dynamic, complex system framework to understand the emergence of mental illnesses—for example, with difficulty returning to homeostasis (critical slowing down) followed by a failure state of the brain.8 Such failure states can become stable within new homeostatic ranges in a process termed allostasis.9 A dynamic, complex system can then remain fixed in a failure state, requiring great energy to dislodge it from its “local minimum.” In fact, a failure state may need to be overcome by paradoxically making the system less resilient so that the system can be reset in a healthy homeostatic range.

Viewed from this angle, studying the rules that govern dynamic complexity in the brain holds promise for understanding the emergence of psychopathology. Contemporary biological psychiatry research has been sterile. Models that posit a specific molecular/cellular/circuit lesion cannot capture the relevant reality because the brain abnormalities in our patients have to do with modulation of physiologic parameters over time, not the functioning of any one brain region or molecule. Current biological psychiatry research can identify factors that are ultimately reflected in the relevant physiologic abnormalities, but they are not by themselves such abnormalities. This probably explains why we repeatedly come up against a one-to-many and many-to-one relationship between biology and clinical presentation—where the same brain abnormalities are seen in multiple disorders, and a single disorder is associated with multiple brain abnormalities.2

Many biological processes contribute to the emergence of psychopathology, and these processes are implicated in many different psychiatric disorders. I believe that no serious worker in the field now expects that we will discover a molecular, cellular, circuit, psychological, behavioral, or other abnormality that by itself will provide meaningful new insights into the emergence of psychiatric disorders, or will explain patterns of comorbidity, heterogeneity, and variability over time. Instead, most advances have served to highlight the complexity of the biology of psychiatric disorders and the absence of a clear way forward through the investigation of any single biological process.10 Studies continue to implicate glutamatergic, GABAergic, and other neurons, as well as glial cells, neuroinflammation, synaptic biology, myelin biology, mitochondrial function, and a host of other processes—each of small effect, and without providing a description of how abnormalities in these systems may lead to symptom formation. This cannot be the whole story since there must be specific types of brain activity that are of critical interest for psychiatry research because they correspond to, or underlie, the abnormal experiences. To be clear, I am not making an etiological argument here; that is, dynamic system abnormalities are not necessarily the cause of psychiatric disorders. Rather, manifestation of psychiatric disorders must involve dynamic system abnormalities as part of the mechanism, and once triggered, these abnormalities may become part of the causal pathway.

New research is needed to capture abnormalities “where the action is”—that is, in dynamically changing patterns of physiologic parameters. Such efforts can help us understand how ongoing brain function has become abnormal and leads to symptom formation in specific presentations. This is easier said than done, because techniques are simply not available to study some of the processes that we would like to study, and because we do not know how to understand and interpret certain kinds of data even if we collected them. There is an urgent need to develop better tools in systems neuroscience.

In this domain, I advocate for designing studies that probe dynamic processes around how the brain self-organizes to maintain homeostasis in the face of disruptive influences, how the brain “tolerates” faulty activity of its components while maintaining integrity of network function, and the molecular/cellular correlates of self-restoration or self-stabilization. These themes are already accessible in animal models, and the BRAIN Initiative (Brain Research Through Advancing Innovative Neurotechnologies) supported by the National Institutes of Health is already making progress on some of them. In human subjects research, certain approaches can be quite useful. Examples include the following: measurements of spontaneous brain oscillations and their response to brain stimulation; stability of, and transitions between, brain states using EEG or resting-state fMRI; and network analysis of changes in symptom presentation in longitudinal studies. With each approach, one can examine oscillations, time constants governing time-series data, and transition patterns. These studies should be longitudinal and coupled with detailed symptom-level clinical information. We are far from being able to identify brain-activity patterns that correspond to specific psychiatric abnormalities, but in each case we can gain insights into how the modulation of ongoing brain function leads to abnormal processing.

Psychiatric neuroscience remains unsatisfactory in its ability to generate mechanistic insights, inform diagnostic systems, and provide stakeholders guidance for prognosis and treatment selection. This is because patient-oriented studies mostly “miss the action” by focusing on a variety of biological levels alone or in combination, and by thereby missing the meta-level abnormalities in dynamic, complex systems that give rise to psychiatric disorders. Informed by our knowledge of brain function, this framework offers an opportunity to identify biological processes that can be mapped onto psychopathology. Ultimately, the hope is to improve our understanding of the patterns of comorbidity, heterogeneity, and variability over time that all clinicians observe. This would provide future trainees, clinicians, patients, and families with a satisfying neuroscientific basis for our field.

Declaration of interest: The author reports no conflicts of interest. The author alone is responsible for the content and writing of the article.

Back to Top | Article Outline


1. Hyman SE. The diagnosis of mental disorders: the problem of reification. Annu Rev Clin Psychol 2010;6:155–79.
2. Paulus MP, Thompson WK. The challenges and opportunities of small effects: the new normal in academic psychiatry. JAMA Psychiatry 2019;76:353–4.
3. Buzsaki G, Draguhn A. Neuronal oscillations in cortical networks. Science 2004;304:1926–9.
4. Aitchison L, Lengyel M. With or without you: predictive coding and Bayesian inference in the brain. Curr Opin Neurobiol 2017;46:219–27.
5. Sterling P. Homeostasis vs allostasis: implications for brain function and mental disorders. JAMA Psychiatry 2014;71:1192–3.
6. Petzschner FH, Weber LAE, Gard T, Stephan KE. Computational psychosomatics and computational psychiatry: toward a joint framework for differential diagnosis. Biol Psychiatry 2017;82:421–30.
7. Redish AD, Kazinka R, Herman AB. Taking an engineer’s view: implications of network analysis for computational psychiatry. Behav Brain Sci 2019;42:e24.
8. Nelson B, McGorry PD, Wichers M, Wigman JT, Hartmann JA. Moving from static to dynamic models of the onset of mental disorder: a review. JAMA Psychiatry 2017;74:528–34.
9. Sterling P, Eyer J. Biological basis of stress-related mortality. Soc Sci Med E 1981;15:3–42.
10. Kendler KS. The dappled nature of causes of psychiatric illness: replacing the organic-functional/hardware-software dichotomy with empirically based pluralism. Mol Psychiatry 2012;17:377–88.
© 2019 President and Fellows of Harvard College