Next-Generation Neurofeedback for Stress-Related Disorders
- John Davis
- Aug 23
- 15 min read
Executive Summary
Stress-related disorders rank among the largest global health burdens, and current combinations of psychotherapy and medication leave many patients without adequate relief.
In a 2024 opinion article in Trends in Neurosciences, Florian Krause, David Linden, and Erno Hermans argue that neurofeedback is uniquely positioned to close the gap between what neuroscience has learned about stress and what clinicians can actually do about it (Krause et al., 2024).
This post traces their reasoning in five steps: the network-level neuroscience of acute stress, the unusual position neurofeedback occupies among treatments, the technical progression from single-region to whole-network training, the two-stage model in which regulation is learned in the clinic and applied in daily life, and the prevention agenda that follows. It then weighs the strengths and limitations of a paper that proposes a research program rather than reporting a trial.
The central message for practitioners is a change in what counts as success. If the authors are right, the outcome that matters is not signal change during a session but whether a client can summon the learned regulation when a real stressor arrives.
What Is the Science?
Stress resists tidy classification because it is confined to neither a single diagnosis nor a single brain region. Stress contributes across anxiety, mood, and trauma-related disorders, and maladaptive biological and psychological responses to stressors can become central to how psychopathology develops and persists.
Krause, Linden, and Hermans (2024) argued that despite substantial progress in understanding the neuroscience of stress, translating that knowledge into effective prevention and treatment has lagged badly behind. They propose neurofeedback as one way to close this translational gap, because it teaches individuals voluntary control over aspects of their own brain activity.
This section lays out the network neuroscience that makes their proposal coherent.
The neuroscience underlying their argument is increasingly network-based. Acute stress produces rapid changes in monoamine neurotransmitters such as norepinephrine, dopamine, and serotonin, followed by slower endocrine responses involving epinephrine and cortisol.
These chemical events do not simply switch individual regions on or off. Instead, they reorganize activity across interacting large-scale brain networks, sets of widely distributed regions that operate together as a system. This distinction matters clinically, because it suggests that a treatment aimed at one structure may be aimed at the wrong level of organization.
Three networks are especially important. The salience network, which includes the dorsal anterior cingulate and dorsomedial prefrontal cortex, anterior insula, temporoparietal junction, thalamus, striatum, amygdala, and hypothalamus, is upregulated during acute stress and coordinates responses to homeostatic threats.
The executive control network supports goal-directed planning, working memory, and endogenous attention, higher-order processes that are typically suspended during acute stress.
The default mode network supports internally directed cognition, including self-referential thought and prospective memory. Healthy adaptation depends partly on the brain’s capacity to switch dynamically among these configurations, and research shows that stress effects are actively reversed in the aftermath of exposure.
This network perspective changes the conceptual target of neurofeedback. Rather than treating anxiety, post-traumatic stress disorder (PTSD), depression, and related conditions as disorders of isolated structures such as the amygdala or prefrontal cortex, Krause et al. (2024) describe stress-related psychopathology as maladapted configurations of large-scale networks and their dynamic interactions. In their terminology, these conditions are brain network disorders, and mitigating them efficiently will likely require network-based interventions. That reframing is the hinge on which the rest of the argument turns.
Neurofeedback interests the authors here because it combines two properties that usually occur separately. It targets brain activity directly, but the regulation comes from endogenous neuromodulation generated by the person rather than applied from outside. Pharmacology and brain stimulation, including transcranial magnetic stimulation, transcranial electrical stimulation, and focused ultrasound, also target the brain directly, but they produce change through an external agent.
Psychotherapy, mindfulness, and behavioral self-regulation depend on processes the individual generates, but they do not specify brain activity itself as the direct treatment target. Neurofeedback occupies the otherwise empty intersection of direct brain targeting and endogenous self-regulation, and the authors build on that position.
Which Ideas Did Krause and Colleagues Present?
Judging this article fairly requires knowing what kind of article it is, so this section describes its design and its four core claims. The paper is an opinion and perspective piece rather than a clinical trial. The authors did not recruit participants, randomly assign patients to conditions, or test a protocol. Instead, they synthesized developments in stress neuroscience, neurofeedback, network neuroscience, ecological assessment, resilience research, and personalized medicine into a conceptual model.
Their central question was broader than whether neurofeedback reduces symptoms. They asked how neurofeedback might be redesigned to match what contemporary neuroscience says about the temporal and neural dynamics of stress-related psychopathology. That shift from whether the method works to what the method should become is the article’s organizing move.
The authors identify four characteristics that make neurofeedback especially promising. It can directly target complex brain network dynamics, learned regulation appears transferable beyond the laboratory, training can occur prospectively before a disorder develops, and both targets and strategies can be adapted to the individual.
Together, these lead to a proposal for a personalized and preventive neuroscience-based intervention focused on regulating stress-related networks when stressors actually occur in everyday life.
The conceptual shift is worth stating plainly. Neurofeedback would no longer be viewed only as a treatment in which a patient normalizes brain activity while connected to equipment. Instead, the clinical sessions become a training environment in which a person acquires a brain-regulation skill that is later deployed independently. In short, the authors reframe neurofeedback from a procedure into a curriculum.
What Was Their Reasoning?
The authors build their case by tracing several generations of neurofeedback technology and then attaching that progression to a two-stage clinical model. Traditional electroencephalography (EEG) neurofeedback typically trains relatively simple parameters, such as power within particular frequency bands or ratios between bands. EEG is comparatively accessible, but its limited spatial resolution makes precise targeting of deep and distributed neural systems difficult. Krause et al. (2024) note that although EEG neurofeedback has shown moderate clinical potential in some stress-related disorders, reported improvements are often difficult to distinguish from nonspecific treatment effects, and overall efficacy remains largely inconclusive.
Newer hemodynamic approaches offer substantially greater anatomical specificity. Functional near-infrared spectroscopy (fNIRS) can be used in real time to target cortical hemodynamic activity, while real-time functional magnetic resonance imaging (rtfMRI) reaches both cortical and deeper structures involved in affective and stress processing. Several controlled studies using these approaches have reported effects on clinical outcomes in anxiety, PTSD, depression, obsessive-compulsive disorder (OCD), fibromyalgia, and obesity.
The authors acknowledge a caveat clinicians should keep in view: these signals are indirect measures of neural activity based on local blood flow and can be difficult to interpret in deep structures.
The next development moves beyond individual brain regions altogether. Functional connectivity neurofeedback (FCNef) trains the strength of interactions between selected regions, and effective connectivity neurofeedback (ECNef) extends this by modeling directional and potentially causal relationships through dynamic causal modeling.
Decoded neurofeedback (DecNef) uses multivariate methods and machine learning to identify and reinforce spatially distributed activity patterns.
Network neurofeedback (NetNef) goes further still by targeting configurations and interactions among entire functional networks, and proof-of-concept studies have already trained regulation of connectivity, differential activation between networks, and even whole functional connectomes.
One implication deserves comment that the authors themselves do not draw. Krause et al. (2024) treat EEG as spatially limited and build their network argument almost entirely on real-time fMRI. Source-localized approaches such as low-resolution electromagnetic tomography (LORETA) neurofeedback, however, are designed precisely to address that spatial objection, and z-score LORETA protocols aim to train activity and connectivity within the executive control, salience, and default mode networks.
If those systems deliver the spatial specificity they target, the authors’ network argument extends to a far more accessible and portable technology than the one they describe. This extension is offered here as clinical commentary rather than as a claim made in the source article.
The authors then connect these technologies to a two-stage intervention model. During the first stage, a person undergoes neurofeedback in a laboratory or treatment setting and learns strategies for voluntarily regulating stress-related brain activity. During the second stage, the individual recalls and applies those strategies without neurofeedback equipment when a meaningful stressor occurs in daily life. The authors call this second stage ecological momentary neuromodulation, and it is the heart of the proposal.
The model depends critically on transfer. Neurofeedback would have limited ecological value if people could regulate their brains only while watching a feedback display. Evidence reviewed by the authors indicates that individuals learn explicit mental regulation strategies and can reproduce regulation after feedback has been removed, a result known as the transfer effect and often treated as a requirement for claiming successful training.
The precise learning mechanisms remain debated, with operant conditioning the most prominent explanation, but transfer itself is established well enough to serve as a foundation. The reasoning therefore runs from network targets, through tools with better spatial resolution, to a skill that outlives the session.
What Were Their Main Conclusions and Implications?
Because this is a perspective article rather than an experiment, there is no single set of statistical findings; instead, this section summarizes the argument’s conclusions and the timing insight that drives them. The principal conclusion is integrative. Advances in neurofeedback and network neuroscience have reached a point where it may be possible to train individuals not merely to alter activity in one brain region, but to regulate distributed neural states relevant to stress and then use that learned regulation at moments of actual need.
The authors find the existing evidence promising but incomplete. Clinical studies have demonstrated potential for neurofeedback targeting individual network nodes in anxiety, PTSD, depression, OCD, and other conditions. Connectivity-based approaches have been applied experimentally in anxiety, depression, and obesity, while proof-of-concept network studies show that people can learn to regulate complex parameters involving connectivity, differential activation between networks, and larger connectome patterns. What has not yet been shown is that any of this prevents illness.
The most innovative part of the article concerns timing. Stress-related symptoms are often episodic, and panic, fear, traumatic intrusions, worry, and urges frequently occur outside scheduled treatment sessions. Because stress-related brain-network changes are themselves transient and time-dependent, the authors argue that interventions should become similarly time-sensitive, a design philosophy known as a just-in-time adaptive intervention.
Neurofeedback could supply a learned skill that is available while the stress response is occurring rather than only when the patient happens to be in a clinic.
This logic also extends neurofeedback from treatment into prevention. People at elevated risk could in principle learn regulation skills before symptoms become clinically significant. The authors note that deterioration in resilience, including slower recovery across repeated stressors, may precede transition into mental illness, and that ecological momentary assessment and wearable ecological physiological assessment may eventually flag these periods of rising vulnerability. Neurofeedback could then be delivered prospectively to strengthen resilience before the transition occurs, an approach the authors suggest could extend even to adolescents and young children.
The same principle supports secondary prevention. Relapse and recurrence are common in stress-related disorders, with rates reaching 50% in some conditions, so neurofeedback added after successful initial therapy might give patients a self-regulation skill that helps them maintain recovery.
Finally, neurofeedback lends itself to personalization. Neural targets can be selected according to an individual’s own functional network organization, and people frequently discover different mental strategies for regulating the same target successfully. Once learned, those strategies belong to the patient rather than to the equipment. This feature can be genuinely empowering because it returns control to individuals in situations where they previously felt they had none.
What Were the Strengths and Limitations?
The article's major strength is its integration of several rapidly developing fields. Instead of treating neurofeedback as an isolated technology, Krause et al. (2024) connect it to network neuroscience, real-time neuroimaging, stress physiology, ecological assessment, resilience research, prevention science, and personalized medicine. This produces a coherent explanation of why neurofeedback might have capabilities that stay invisible if one only asks whether a conventional protocol reduces symptoms. It also leaves the field with testable predictions rather than a general endorsement.
A second strength is the move from regional to network-level targets. If stress-related psychopathology reflects disturbed interactions among salience, executive control, default mode, and other systems, then training only one anatomical structure may be an incomplete intervention. Network neurofeedback provides a conceptual way to match the treatment target to the distributed nature of the pathology.
A third strength is the emphasis on ecological validity. The authors directly confront a persistent problem in clinical neuroscience: showing a signal can change inside a scanner does not establish that the change will matter when a person meets a real-world stressor. Their two-stage model makes transfer into daily life a central therapeutic requirement rather than an optional secondary outcome. They also note that neurofeedback is generally safe, with only a few mild and transient side effects reported, which matters for any intervention proposed for people who are not yet ill.
The limitations are equally important. The article presents a hypothesis-driven framework, not proof that the proposed preventive intervention works. The authors explicitly identify unanswered questions about how long learned regulation lasts, whether people can reliably use it during actual stressful situations, whether real-world application changes stress and resilience, whether prospective training prevents psychopathology, and whether targeting whole networks produces greater clinical benefit than targeting individual nodes.
The transfer problem deserves particular scrutiny. Although transfer after neurofeedback has been demonstrated, most studies have assessed it relatively soon after training, so long-term retention needs direct investigation and refresher sessions may prove necessary. The preventive claims ultimately require randomized controlled trials measuring whether prospective neurofeedback reduces the later emergence of clinical symptoms. The authors identify PTSD as an attractive first test case, because trauma provides an identifiable at-risk moment and intrusive memories provide recognizable moments for real-world intervention.
Individual learning ability presents another limitation.
Roughly 15% to 30% of individuals across studies have difficulty learning regulation of the targeted neural parameter, a phenomenon often described as BCI illiteracy. Predicting who will respond to neurofeedback could therefore become an important component of personalized treatment, and the authors also raise the possibility of supporting nonresponders with concurrent brain stimulation during initial training.
Cost and access are the final constraints. Sophisticated neurofeedback, particularly fMRI-based neurofeedback, is expensive and difficult to scale, while EEG and fNIRS are more accessible but introduce trade-offs in spatial resolution and target selection.
The authors envision future systems in which mobile neuroimaging, peripheral biofeedback, wearable sensors, and smart devices complement laboratory-based training. Whether source-localized EEG can close enough of the spatial gap to make network training routinely affordable is, in this writer’s view, the most consequential open question for practicing clinicians.
What Was the Impact?
The proposal's conceptual impact is that it reframes what successful neurofeedback for stress-related disorders could ultimately look like. The goal would not simply be to produce a change in EEG or fMRI activity during a training session. The clinically meaningful endpoint would be the person’s ability to use the learned regulation when stress occurs. Neurofeedback therefore becomes a form of skill acquisition, with the neuroimaging system functioning as a training instrument rather than a permanent part of the long-term intervention.
This approach also changes the temporal orientation of treatment. Conventional mental health care is largely reactive, in that symptoms develop, a diagnosis is made, and treatment begins. The proposed neurofeedback model is partly prospective, training people known to be at elevated risk before a critical transition occurs. The ultimate goal is therefore not merely symptom reduction but resilience, relapse prevention, and maintenance of mental health.
The article’s central figure crystallizes this argument. It places neurofeedback at the intersection of direct brain targeting and endogenous intervention, then depicts neurofeedback training followed by ecological momentary neuromodulation in response to a real-life stressor. The final panel illustrates the preventive hypothesis, showing how repeated stressors may progressively erode mental health while prospective neurofeedback creates opportunities to intervene at vulnerable moments and alter that trajectory.
The importance of this paper therefore lies less in demonstrating that neurofeedback has already solved stress-related disorders than in defining a research agenda for what the next generation of neurofeedback might become. The authors envision a progression from simple brain signals to networks, from clinic-based regulation to real-world application, from generic protocols to individualized targets, and from treating established illness to strengthening resilience before or after illness occurs.
Their conclusion is deliberately forward-looking, positioning neurofeedback as a personalized preventive intervention capable of producing ecological momentary neuromodulation when real stressors occur. For clinicians, the practical message is to begin treating transfer, rather than in-session signal change, as the outcome worth measuring.
Takeaways for Stress-Related Disorders
1. Stress-related disorders are increasingly understood as disorders of dynamic brain networks. Acute stress changes interactions among the salience, executive control, and default mode networks rather than affecting isolated brain regions.
2. Neurofeedback occupies an unusual therapeutic position because it is both direct and endogenous. It targets brain activity directly while teaching individuals to generate the regulatory change themselves.
3. The next generation of neurofeedback moves from single regions toward connectivity and whole-network regulation. Functional connectivity, effective connectivity, decoded, and network neurofeedback offer increasingly sophisticated ways of targeting distributed neural dysfunction.
4. Transfer into everyday life is central to the proposed clinical model. Regulation is first learned with feedback and later deployed without equipment during real-world stress, a process the authors call ecological momentary neuromodulation.
5. The most ambitious implication is prevention rather than treatment alone. Prospective training could strengthen resilience and reduce relapse, but this remains a hypothesis requiring longitudinal research and randomized controlled trials.
Glossary
BCI illiteracy: the observation that a minority of individuals, commonly estimated at 15% to 30% across studies, do not learn to regulate the targeted neural parameter during brain-computer interface training.
brain-computer interface (BCI): a technological system that uses measured brain activity to interact with or control a computer or another external device and that provides the closed-loop infrastructure required for neurofeedback.
brain network disorder: a mental disorder conceptualized as involving maladaptive organization or interaction of distributed large-scale neural networks rather than dysfunction confined to a single brain region.
decoded neurofeedback (DecNef): a neurofeedback approach that uses multivariate analysis and machine learning to identify and train spatially distributed patterns of brain activity.
default mode network: a large-scale brain network involving medial prefrontal, posterior cingulate, precuneus, parietal, and medial temporal regions that supports internally directed cognition, including self-referential and prospective mnemonic processing.
ecological momentary assessment (EMA): a longitudinal methodology that repeatedly samples an individual’s current experiences or behaviors in everyday environments, frequently through mobile devices.
ecological momentary neuromodulation: the proposed application of previously learned endogenous brain-regulation strategies at relevant moments in daily life, particularly following exposure to actual stressors.
ecological physiological assessment: a longitudinal approach that repeatedly measures physiological processes in everyday life, often using wearable devices such as smartwatches.
effective connectivity neurofeedback (ECNef): a form of neurofeedback that uses models such as dynamic causal modeling to train directional interactions between neural regions rather than connectivity strength alone.
electroencephalography (EEG): an electrophysiological technique that measures electrical potentials recorded from the scalp and is widely used to provide signals for neurofeedback.
endogenous neuromodulation: alteration of neural activity generated internally by the individual through learned mental regulation rather than through externally administered drugs or stimulation.
executive control network: a large-scale neural system involving dorsolateral and dorsomedial prefrontal cortex, precentral and superior frontal sulci, and posterior parietal areas that supports goal-directed planning, working memory, endogenous attention, and other higher-order cognitive functions.
functional connectivity: a statistical relationship between activity occurring in different brain regions that can be used to characterize communication or coordinated activity within neural systems.
functional connectivity neurofeedback (FCNef): a neurofeedback method designed to train individuals to modify the strength of functional connectivity between selected brain regions.
functional magnetic resonance imaging (fMRI): a neuroimaging technique that infers brain activity from hemodynamic changes and permits relatively high spatial resolution throughout the brain.
functional near-infrared spectroscopy (fNIRS): an optical neuroimaging technique that measures hemodynamic changes near the cortical surface using near-infrared light.
just-in-time adaptive intervention (JITAI): a real-world intervention designed to deliver appropriate support when it is most needed by adapting to changes in an individual’s internal state and environmental context.
large-scale brain network: a distributed collection of interconnected brain regions that function together as a neural system and can often be identified through functional connectivity analyses.
low-resolution electromagnetic tomography (LORETA): a source-localization method that estimates the intracerebral generators of scalp-recorded electrical activity and permits electroencephalographic training of deeper and distributed targets.
network neurofeedback (NetNef): a neurofeedback approach that trains regulation of entire large-scale brain networks or interactions among networks rather than activity in a single brain region.
neurofeedback: a form of biofeedback in which information about a selected neural signal is presented to an individual in real time so that voluntary regulation of that neural parameter can be learned.
neuromodulation: the alteration of neural activity or neural-system dynamics through endogenous strategies or exogenous interventions such as pharmacology or brain stimulation.
operant conditioning: a learning mechanism in which behavior is modified by its consequences and one of the principal theoretical mechanisms proposed to explain acquisition of neural self-regulation during neurofeedback.
real-time functional magnetic resonance imaging (rtfMRI): a form of fMRI in which brain data are processed and analyzed as they are acquired, permitting the resulting signal to be incorporated into a brain-computer interface and used for neurofeedback.
resilience: an individual’s capacity to adapt successfully to stressor exposure and recover without progressing toward persistent maladaptive responses or psychopathology.
salience network: a large-scale brain system involving the anterior insula, dorsal anterior cingulate and dorsomedial prefrontal cortex, temporoparietal junction, thalamus, striatum, amygdala, and hypothalamus that helps identify biologically significant events and coordinate appropriate cognitive, behavioral, autonomic, and neuroendocrine responses.
stress-related disorder: a mental disorder characterized by maladaptive biological and psychological responses to short-term or long-term exposure to physical or emotional stressors.
stressor: a physical or psychological event that threatens an organism’s homeostasis.
transfer effect: the ability to reproduce a neural self-regulation skill learned during neurofeedback after the feedback itself has been removed.
References
Krause, F., Linden, D. E. J., & Hermans, E. J. (2024). Getting stress-related disorders under control: The untapped potential of neurofeedback. Trends in Neurosciences, 47(10), 766–776. https://doi.org/10.1016/j.tins.2024.08.007
About the Author
Dr. John “Dusty” Davis has devoted his career as a neuropsychologist to clinical services and research in both hospital and community settings with people who have experienced brain injury of various types. His academic career is based at the Department of Psychiatry and Behavioural Neurosciences at McMaster University in Hamilton, Ontario Canada. Having been registered with the Ontario College of Psychologists and Behaviour Analysts and certified in behavioral and cognitive psychology by the American Board of Professional Psychology for many years, he is also recognized by the Biofeedback Certification International Alliance as qualified in neurofeedback.
Support Our Friends
Comments