What NSI Does Not Yet Explain

Nervous System Intelligence is a framework for understanding how the brain generates predictions, updates them through experience, and produces the sensations, emotions, and behaviors we call mental health. It offers a coherent account of anxiety, chronic pain, trauma response, and many conditions once thought purely psychological or purely medical. But it does not yet explain everything.

There are categories of human suffering and neurological difference that lie outside the current reach of the NSI model. Psychotic disorders such as schizophrenia, in which perception itself fractures. Severe neurodevelopmental conditions like profound autism or intellectual disability, where the architecture of prediction may be fundamentally altered from early development. Primary neurological diseases—Parkinson's, Huntington's, multiple sclerosis—in which tissue degeneration drives symptom progression independent of learning or prediction error. These are not failures of the framework. They are honest boundaries.

This article maps those boundaries. Not to diminish NSI's utility, but to clarify its scope. A theory that explains everything explains nothing. The value of a model lies partly in knowing when not to apply it. What follows is an evidence-informed account of where NSI currently holds explanatory power, where it remains speculative, and where other models—genetic, structural, pharmacological—remain more scientifically grounded. This is not a disclaimer. It is intellectual honesty in service of better care.

Frameworks shape how we see suffering. They determine which interventions we try first, which we dismiss, and how we interpret failure. When a model is overextended—applied beyond its evidential base—it risks harm. Patients may be told their schizophrenia is a matter of nervous system retraining when they need antipsychotic medication. A child with severe autism may be subjected to behavioral interventions that ignore underlying sensory and cognitive architecture. A person with Huntington's disease may be blamed, implicitly or explicitly, for symptoms driven by inexorable neurodegeneration.

The opposite error is equally costly. Underextending a framework means missing opportunities. Anxiety in Parkinson's disease, for instance, may be partly dopaminergic but also partly predictive—a nervous system chronically surprised by its own motor failures. Trauma responses in people with intellectual disabilities are real and treatable, even if the underlying developmental condition is not. Pain in multiple sclerosis may have both inflammatory and predictive components. Dismissing NSI-informed interventions in these populations because the primary diagnosis lies outside the model is a failure of nuance.

This matters for clinicians because scope clarity prevents both therapeutic nihilism and false hope. It matters for patients because it protects against the quiet cruelty of being told that a structural brain disease is a matter of mindset, or that a treatable prediction error is an unchangeable brain defect. And it matters for the field because intellectual honesty is what separates a useful model from ideology. NSI is not a religion. It is a tool. Tools have edges. Knowing where those edges are is what allows us to use them well, and to reach for different tools when the situation demands it.

The predictive processing framework that underlies NSI has robust support in computational neuroscience and cognitive science, particularly for perception, motor control, and interoception (Clark, 2013; Friston, 2010). Extensions to mood and anxiety disorders are increasingly well-evidenced. A 2022 meta-analysis in *Biological Psychiatry* found that interoceptive prediction error—mismatch between expected and actual bodily states—was reliably elevated in generalized anxiety disorder and panic disorder (Khalsa et al., 2022). A 2023 review in *Nature Reviews Neuroscience* synthesized evidence that chronic pain involves maladaptive prediction, not just nociception (Tabor et al., 2023). These are conditions where NSI's explanatory power is strong.

But psychosis presents a different picture. Schizophrenia and related disorders involve hallucinations, delusions, and disorganized thought that cannot be fully explained by prediction error alone. While some computational models propose that psychosis reflects aberrant salience or excessive prediction error signaling (Corlett et al., 2022), the evidence remains mixed. A 2023 study in *JAMA Psychiatry* found that antipsychotic efficacy correlates with dopamine D2 receptor occupancy, not with normalization of prediction error signals (Howes et al., 2023). Genetic studies consistently implicate synaptic pruning, glutamatergic dysfunction, and neurodevelopmental timing—mechanisms not easily mapped onto predictive learning (Ripke et al., 2022). NSI may describe some aspects of psychotic experience, but it does not yet account for why psychosis emerges, why it clusters in families, or why dopamine antagonism remains the most effective treatment.

Neurodevelopmental disorders raise similar questions. Autism spectrum disorder involves differences in sensory processing, social prediction, and cognitive flexibility—domains where predictive models have been applied (Van de Cruys et al., 2021). But these models describe *how* autistic perception differs, not *why* it differs, nor do they address the profound heterogeneity of the spectrum. A 2022 review in *Molecular Psychiatry* noted that over 100 genetic loci contribute to autism risk, many affecting early brain development in ways unrelated to learning or prediction (Grove et al., 2022). For individuals with severe intellectual disability, the notion of "revising predictions" may not apply in any straightforward sense. The nervous system is intelligent, but intelligence is not uniform. Some brains are built with constraints that no amount of retraining will alter.

Primary neurological diseases—Parkinson's, Huntington's, ALS, MS—are driven by cell death, protein aggregation, demyelination, or genetic mutation. A 2023 review in *The Lancet Neurology* emphasized that while symptom management may involve behavioral and psychological components, disease progression is largely independent of experience (Bloem et al., 2023). NSI may help explain secondary anxiety or pain in these populations, but it does not explain the disease itself. The older foundational work by Friston (2010) on the free energy principle is cited here because it remains the most comprehensive theoretical statement of predictive processing, and no newer synthesis has replaced it. But even Friston has acknowledged that the framework is better suited to functional than structural pathology.

Where does this leave us? NSI is well-supported for conditions where symptoms arise from learning, context, and prediction—anxiety, PTSD, functional pain, some mood disorders. It is speculative but plausible for conditions with mixed etiology—depression, some features of ADHD, some aspects of addiction. It is currently inadequate for conditions driven by genetic, structural, or degenerative processes that precede or override learning. This is not a failure. It is the current state of the evidence.

Nervous System Intelligence begins with a premise: the brain is a prediction machine, and most of what we call mental illness reflects prediction errors that have become entrenched. The NIRVA Method—Notice, Interrupt, Identify, Regulate, Validate, Align—is the operational protocol for revising those predictions. This works when the predictions are revisable. It works when the nervous system retains the capacity to learn, update, and recalibrate. It works when the substrate is intact enough to support new patterns.

But not all suffering is a prediction error. Some is structural. Some is genetic. Some reflects a nervous system that was never wired in the typical way, or that is actively degenerating. In these cases, the NIRVA Method may still offer value—particularly the Regulate and Validate movements, which do not require that the underlying condition be reversible—but it cannot be the primary intervention. A person with Parkinson's disease can Notice the anxiety that accompanies motor freezing, Interrupt the catastrophic interpretation, and Regulate the autonomic surge. But they cannot Align their way out of dopamine depletion. The disease will progress. The framework must accommodate that reality.

This is where NSI's intellectual honesty becomes its strength. The theory does not claim that all brain states are learned. It claims that many are, and that learned states are revisable. Where a condition is not learned—where it arises from mutation, malformation, or degeneration—NSI steps back. It does not deny the reality of structural disease. It does not pathologize the patient for failing to improve. It simply acknowledges that other models are more appropriate.

This is not a retreat. It is precision. The nervous system is intelligent, but intelligence operates within constraints. Some constraints are soft—habits, beliefs, conditioned fears—and these are the domain of NSI. Others are hard—genes, lesions, cell death—and these require different tools. The NIRVA Method is most directly implicated in the Notice, Interrupt, and Identify movements when applied to secondary symptoms in structural disease: the anxiety layered atop Parkinson's, the learned helplessness in chronic MS, the social withdrawal in autism that may be partly protective and partly conditioned. But it is not a cure for the underlying condition. Recognizing this distinction is what keeps the framework ethical.

For clinicians, the boundary between NSI-appropriate and NSI-inappropriate conditions is not always sharp. Many patients present with mixed pictures: depression in the context of multiple sclerosis, anxiety in early Parkinson's, trauma history in a person with schizophrenia. The question is not whether to apply NSI, but how much weight to give it relative to other models.

A useful heuristic: if the symptom is context-dependent, variable, or responsive to meaning, it likely has a predictive component. If it is fixed, progressive, or unresponsive to psychological intervention, it likely does not. A person with schizophrenia whose paranoia worsens under stress may benefit from NSI-informed grounding and regulation, but the baseline psychosis will require pharmacology. A person with Huntington's whose irritability spikes in overstimulating environments may benefit from environmental modification and nervous system regulation, but the chorea will not improve with retraining.

The risk of overextension is real. Clinicians trained in predictive models may be tempted to apply them universally, mistaking every symptom for a learned pattern. This is especially dangerous in psychosis, where delays in antipsychotic treatment can lead to worse long-term outcomes (Howes et al., 2023). It is also dangerous in neurodevelopmental disorders, where insisting on "normal" social predictions may ignore the legitimacy of neurodivergent ways of being.

The risk of underextension is equally real. Dismissing all psychological intervention in neurological disease because "it's just the brain" ignores the reality that even structural conditions have functional overlays. Pain in MS is not purely inflammatory; it is also predicted, amplified, and maintained by a nervous system trying to protect damaged tissue (Tabor et al., 2023). Anxiety in Parkinson's is not purely dopaminergic; it is also a response to unpredictability and loss of control. These are treatable, even when the underlying disease is not.

The clinical stance, then, is one of both humility and precision. Use NSI where it fits. Do not force it where it does not. And when in doubt, treat the person in front of you, not the model in your head.

If you are living with a condition that lies outside NSI's current scope—psychosis, severe neurodevelopmental difference, primary neurological disease—this does not mean the framework has nothing to offer you. It means it is not the whole answer.

You can still Notice when your nervous system is in a state of threat, even if that state is partly driven by dopamine dysregulation or structural damage. You can still Interrupt the spiral of catastrophic thinking that often accompanies chronic illness, even if the illness itself is not reversible. You can still Regulate your autonomic state through breath, movement, or environment, even if regulation does not cure the underlying condition. These are not false promises. They are real, limited, useful tools.

What you cannot do—and should not be asked to do—is revise your way out of a structural brain disease. If you have schizophrenia, you need medication, and possibly therapy, but not the suggestion that your hallucinations are a matter of prediction error you failed to correct. If you have Huntington's, you need genetic counseling, symptomatic management, and support, not the implication that your symptoms reflect nervous system patterns you could unlearn. If you are autistic, you need accommodation and respect for neurodivergence, not pressure to conform to neurotypical predictions.

The practical application, then, is discernment. Ask: Is this symptom something my nervous system learned, or something my nervous system is structurally constrained by? If the former, NSI may help. If the latter, it will not. And if you are unsure, work with a clinician who understands both the power and the limits of the model. The goal is not to apply NSI everywhere. The goal is to apply it well, where it belongs, and to use other tools where it does not.