Brain Inflexibility Spans 11 Psychiatric Disorders; Rutgers Study Finds Symptoms Outpredict Labels

Three days ago, Rutgers University neuroscientist Michael W. Cole published his new book on brain network dynamics arguing that the movement of activity through brain networks is the foundational mechanism of cognition, creativity, and mental health. The timing is not incidental. Across the same campus, researchers in his colleague’s lab have produced the strongest peer-reviewed evidence yet for that exact thesis.

A study published in Nature Communications by a Rutgers and Yale team found that reduced brain network flexibility — the brain’s ability to reorganize its communication patterns as mental demands change — is a universal biological marker shared across all 11 psychiatric conditions examined. The finding was not confined to one or two diagnoses. It appeared in participants with ADHD, depression, schizophrenia, OCD, PTSD, bipolar disorder, eating disorders, anxiety, panic disorder, substance use disorder, and more. The same brain, losing the same flexibility.

More provocatively: when machine learning models were used to predict these brain patterns, a person’s individualized "symptom fingerprint" — capturing their actual constellation of difficulties across internalizing, externalizing, cognitive, and social-reward dimensions — outperformed their diagnostic label as a predictor of neural biology. The brain, it seems, cares less about the box on the DSM checklist than about how the person is actually suffering.

A Brain That Can’t Shift Gears

The human brain is not a static organ. Its large-scale networks — clusters of regions that coordinate together — continuously reconfigure their connectivity patterns depending on what a person is doing: resting quietly, solving a problem, navigating a charged social moment. This constant reshuffling, what neuroscientists call dynamic network reconfiguration, is a hallmark of healthy cognition. It allows the brain to recruit the right circuitry for each task and release it when no longer needed.

In people with psychiatric disorders, that machinery appears to seize up.

The Rutgers-Yale team found that individuals with psychiatric diagnoses showed markedly reduced flexibility in how their brain networks reorganized across cognitive states compared to healthy controls. Their functional connectivity patterns remained comparatively rigid — less able to differentiate between rest and task demands. The researchers describe this as a "flattening" of brain dynamics: the brain’s repertoire of network configurations is compressed, with less variation between states than in healthy individuals.

Think of it as a city’s road network that cannot adapt to rush hour. The roads are present, the traffic moves — but the patterns don’t shift with conditions. Everything sluggish, regardless of what the moment demands.

The Study: 219 Participants, 11 Diagnoses, Six Cognitive States

The research draws on the Transdiagnostic Connectome Project, a dataset collected at imaging centers at Yale University and McLean Hospital in Belmont, Massachusetts between November 2019 and March 2023. The final analyzed cohort comprised 219 adults — 134 spanning 11 psychiatric diagnoses and 85 healthy controls — with a mean age of 36.2 years.

The diagnostic breadth was deliberate. By including 11 distinct conditions, the researchers could ask whether any brain-dynamics signal transcended diagnostic boundaries rather than merely characterizing a single disorder.

Each participant underwent MRI scanning across six conditions: three resting-state scans, two variants of the Stroop color-word task (a classic test of cognitive control), and an emotional processing task. This generated data across six distinct cognitive states, allowing the team to track how brain network configurations shifted — or failed to shift — as mental demands changed. The researchers also collected 110 behavioral, cognitive, and clinical measurements per participant.

How the measurement works: For each of the six cognitive states, pairwise correlations between 434 brain regions organized into 21 large-scale networks were computed. A computational technique called non-negative matrix factorization (NMF) then decomposed these connectivity matrices to identify fundamental "subgraph" patterns — essentially the brain’s basis vocabulary of connectivity states. The key measure was how similar these subgraph expressions were across states: high similarity meant states looked interchangeable (rigid, flattened); low similarity meant the brain was genuinely reconfiguring (flexible). Statistical results were striking: the case-control difference in network flexibility reached Cohen’s D = 3.3, while the symptom fingerprint prediction outperformed diagnostic category and case-control status (D = 9.6 for diagnosis, D = 17.8–61.2 for symptom fingerprints, depending on measurement approach).

The study was led by Dr. Carrisa V. Cocuzza, a postdoctoral fellow in the Holmes Lab, and Dr. Avram J. Holmes, associate professor of psychiatry at Rutgers Robert Wood Johnson Medical School and core faculty member of the Center for Advanced Human Brain Imaging Research within the Rutgers Brain Health Institute.

Symptoms, Not Diagnoses, Track the Brain Signal

The flattening finding alone is notable. The study’s second major discovery may be even more consequential for the field.

Using hierarchical clustering of their 110 clinical and behavioral measures, the team constructed individualized symptom fingerprints — personalized profiles capturing how each participant scored across four core dimensions of psychiatric functioning: internalizing symptoms such as depression and anxiety, externalizing behaviors, cognitive difficulties, and social and reward processing. These yielded 16 distinct fingerprint clusters across the full cohort.

When machine learning models tested which variable best predicted each participant’s brain network flexibility profile, symptom fingerprints outperformed both traditional diagnostic categories and simple case-control status. The brain cared less about whether someone had been labeled "PTSD" or "bipolar disorder," and more about the specific constellation of symptoms and functional impairments they were actually living with.

"You can learn a lot from traditional symptom measures by isolating a specific aspect of behavior that shifts up and down, which clinicians can also use to track how well treatments are working," said Dr. Cocuzza. "We were interested in the threads that span a diversity of psychiatric symptoms. Our symptom fingerprinting approach builds on latent dimensional models and captures how multiple behaviors vary at once, which reflects how people function in everyday life."

The frontoparietal control network — a set of prefrontal and parietal regions heavily implicated in executive function and cognitive control — emerged as especially central to the observed differences. Through systematic model comparison, the team found that inhibitory cognitive control and frontoparietal network interactions drove most of the gap between patients and healthy controls. This is consistent with existing literature: disruption of the frontoparietal network has been documented across virtually every studied psychiatric and neurological condition. That it appears here as a common denominator of flexibility loss strengthens the case for its role as a broad biological lever in mental illness.

The conditions showing the most pronounced flattening: panic disorder, eating disorder, OCD, and ADHD.

Challenging the Diagnostic Category Model

The broader implication lands squarely in one of psychiatry’s most active debates: whether the categorical diagnostic system enshrined in the DSM — Diagnostic and Statistical Manual of Mental Disorders — is the right framework for understanding and treating mental illness.

This study adds to a growing body of evidence supporting dimensional, transdiagnostic frameworks like the NIMH Research Domain Criteria (RDoC), which seeks to map mental health along spectra of biological and behavioral traits rather than sorting patients into categorical boxes. NIMH launched RDoC in 2010 precisely because DSM categories, while highly reliable, have limited biological validity — the same diagnosis can describe vastly different underlying neural profiles. This study offers direct empirical support for RDoC’s core hypothesis.

Dr. Emily Ferenczi, a neurologist and faculty member at Harvard Medical School, offered independent perspective. "Many recent studies have suggested that dynamics of brain networks are altered in psychiatric disorders — for example, through changes in strength or temporal structure of brain network activity," she noted. "However, most studies to date have focused on a static state, by analyzing how the activity of brain networks is altered in a single resting state in individuals with psychiatric disease."

This study breaks from that pattern by specifically measuring transitions between states, capturing the dynamic dimension most prior work had missed. "They find that the dynamics of brain networks are less variable — ‘flattened’ — when individuals move from one state to another, and that the extent of this flattening can predict diagnosis and symptom profile," Ferenczi said. "The flexible recruitment of different patterns of brain network activity when switching between cognitive states may be altered in psychiatric disease."

She also highlighted the precision medicine implications: "The current study adds a neural circuit dimension to this transdiagnostic approach by demonstrating that brain network dynamics could better classify individuals in terms of their specific symptom fingerprints than by their diagnostic label. This implies that an individual’s unique symptom profile may be a better indicator of the underlying brain network pathology, and thus may allow us to tailor treatment more precisely and specifically to the underlying biological process."

That vision is gaining institutional momentum. The Stanford Commission on Precision Mental Health, led by neuroscientist Dr. Leanne Williams, is separately building frameworks for redefining mental disorders via brain circuit imaging. And this week, Singapore researchers at the National University Hospital and National University of Singapore reported that individually targeting treatment-resistant OCD patients’ brain circuits with personalized OCD brain stimulation guided by each patient’s own imaging data achieved 78% symptom reduction in a pilot patient — a clinical demonstration of the network-targeting principle the Rutgers-Yale study characterizes.

A Warning the Study Itself Doesn’t Name: The Insurance Problem

The science is pointing toward a future where brain-based, symptom-dimensional profiles guide treatment. But there is a structural obstacle this research raises that tends to go unaddressed in psychiatry’s mainstream discourse: the entire economic architecture of psychiatric care runs on DSM codes.

Insurance reimbursement, billing, clinical trial enrollment, and therapeutic authorization all depend on categorical diagnoses. The American Psychiatric Association’s own Future DSM Strategic Committee acknowledged this tension in an early 2026 series of papers, noting in its Future DSM Strategic Committee commentary that integrating biological discoveries into future diagnostic frameworks risks what the committee’s written papers described as a "Rett syndrome problem": when that disorder’s genetic basis was discovered and it was removed from the DSM, some insurers stopped covering related psychiatric treatment — because the diagnostic code was what triggered reimbursement, not the underlying biology. Better science, inadvertently narrowing access.

This is not a reason to slow the science. It is a reason for policymakers, payers, and professional organizations to act preemptively — building reimbursement infrastructure that can accommodate dimensional, biology-informed profiles before clinical tools based on them arrive. The science from Rutgers and Yale, and from Stanford’s commission, suggests that arrival may be closer than insurers are prepared for.

Limitations: Science on the Frontier

The authors and independent commentators were frank about the study’s constraints.

The sample, while transdiagnostically broad, comprised 219 participants — substantial for this type of neuroimaging research, but modest relative to the claims being made. Replication in larger and more diverse cohorts is essential before clinical translation. The study was also collected at two sites (Yale and McLean), which introduces scanner-site variability that must be accounted for. And like virtually all studies in this population, it cannot fully control for medication effects on brain dynamics.

There is also the perennial challenge of moving from group-level statistical patterns to clinically actionable, individual-level predictions. As Dr. Ferenczi noted, "It remains a challenge to apply these computational and statistical approaches in a clinical setting at the level of individual patients."

Most fundamentally, the study is observational. It establishes correlation, not causation. The researchers identify that reduced network flexibility accompanies psychiatric illness and tracks symptom severity — but cannot definitively prove the flattening causes the symptoms. Ferenczi pointed toward a path: "There is an opportunity to perhaps demonstrate causality in the opposite direction through therapeutic approaches that correct the underlying brain network dysfunction, thereby hopefully leading to a causal positive therapeutic benefit" — including neuromodulation research such as transcranial magnetic stimulation and deep brain stimulation as potential avenues for testing causal direction.

---

Frequently Asked Questions

What does "flattened brain dynamics" mean, exactly?

In a healthy brain, the patterns of communication between brain regions change substantially depending on what you’re doing — resting, problem-solving, processing an emotion. "Flattening" means those patterns stay similar across different mental states, as if the brain’s connectivity is stuck in one configuration regardless of what’s demanded of it. The Rutgers-Yale study measured this by tracking how six different cognitive states compared in terms of brain connectivity patterns, finding psychiatric patients showed significantly less differentiation between states.

Can a brain scan now diagnose my mental illness?

No. This research is early-stage and conducted at the group level. No brain-based biomarker for psychiatric diagnosis is currently FDA-approved or in routine clinical use. The study identifies a pattern that holds across groups of patients, but individual-level prediction — accurate enough to guide clinical decisions for a specific person — remains a research goal, not a current clinical tool.

What is the DSM, and why does this study challenge it?

The Diagnostic and Statistical Manual of Mental Disorders (DSM) is the American Psychiatric Association’s classification system for mental health conditions, now in its fifth edition (DSM-5). It organizes psychiatric illness into discrete categories based on symptom checklists. Critics note that these categories have high clinical reliability — different clinicians using the DSM reach consistent diagnoses — but limited biological validity, meaning they don’t map cleanly onto distinct biological processes in the brain. This study adds to evidence suggesting that brain biology aligns more closely with dimensional symptom profiles than with categorical labels.

What is RDoC, and what does this study mean for it?

Research Domain Criteria (RDoC) is an NIMH framework launched in 2010 to support research that maps mental health along biological dimensions rather than DSM categories. The Rutgers-Yale finding that symptom fingerprints outperform diagnostic labels in predicting neural patterns is direct empirical support for RDoC’s premise — that biology-informed, dimensional measures will ultimately prove more scientifically valid than categorical diagnosis.

What comes next for this research?

The Rutgers team plans to study how brain network dynamics change over time and in response to interventions. The parallel development of personalized brain stimulation therapies — such as this week’s NUH/NUS report of 78% OCD symptom reduction using individually targeted transcranial magnetic stimulation — suggests the field is moving toward testing whether restoring brain flexibility is itself a therapeutic goal, not just a diagnostic signal.

The Road Ahead

"The findings indicate that impairments in time-varying brain processes may underlie differences in cognitive and behavioral functioning across individuals," said Dr. Holmes. "By showing that brain network dynamics track symptom variation across diagnostic boundaries, the study supports a shift toward more personalized, biology-informed models of mental health."

If the findings hold and replicate, the implications extend well beyond academic neuroscience. Future psychiatric treatment — whether pharmacological, behavioral, or neuromodulatory — might be optimized not by targeting disease categories but by targeting a person’s specific profile of network rigidity. Treatment calibrated to where along the internalizing, externalizing, cognitive, and social-reward dimensions a person struggles most. Monitored by tracking whether their brain dynamics have regained flexibility.

That is a vision of psychiatry that looks quite different from today’s. But the science now points toward it from multiple directions: from fMRI studies at Rutgers and Yale, from precision medicine commissions at Stanford, from Cole’s new book arguing that brain flows are the substrate of mind — and from a clinical tradition of therapeutic work that is beginning to target those flows directly.

The brain hasn’t forgotten how to shift. The scientific case for helping it remember has never been stronger.

The original paper, "Brain network dynamics reflect psychiatric illness status and transdiagnostic symptom profiles across health and disease," by Carrisa V. Cocuzza, Sidhant Chopra, Ashlea Segal, Loïc Labache, Rowena Chin, Kaley Joss, and Avram J. Holmes, was published July 21, 2026, in Nature Communications open access article (Vol. 17, Article No. 6678). The study is open access. The Transdiagnostic Connectome Project dataset is publicly available at OpenNeuro dataset ds005237.