The schizophrenia story that is emerging is complex and fascinating. Unraveling the interplay between neural circuits and molecular mechanisms in the in the PFC is just beginning.
David Lewis has done the best research on the glutamate hypothesis. Ketamine is an NMDA receptor blocker, which triggers the downregulation of GAD67 in PV basket cells and somatostatin interneurons. The story is quite complex with interleukin 6 as an intermediate between them.
Dienel, S. J., Fish, K. N., & Lewis, D. A. (2023). The nature of prefrontal cortical GABA neuron alterations in schizophrenia: markedly lower somatostatin and parvalbumin gene expression without missing neurons. American Journal of Psychiatry, 180(7), 495-507.
Dr. Sejnowski, this is an excellent and important contribution. I particularly appreciate the move away from viewing schizophrenia primarily through the dopamine hypothesis and toward a model involving inhibitory interneurons, gamma coordination, and network timing. Your findings concerning GAD67 reduction and parvalbumin basket-cell dysfunction provide a plausible biological mechanism by which cortical activity could become both hyperactive and poorly synchronized.
In my own theoretical work on schizophrenia as a disorder of conscious moment formation, I have suspected that positive symptoms may be associated especially with posterior cingulate cortex and default-mode network dysfunction. The PCC contributes contextual integration, self-referential processing, and construction of an internal model of reality; impaired inhibitory timing in these networks might therefore allow internally generated representations to compete abnormally with sensory reality. I have also wondered whether the brain may compensate for this posterior instability by suppressing broader network propagation, thereby reducing hallucinations at the cost of producing secondary negative symptoms such as avolition, flattened affect, and diminished behavioral initiation. Have you encountered physiological or imaging evidence supporting either of these anterior-posterior or compensatory trends?
I discuss this possibility more fully in my Substack article, “Schizophrenia as a Network Disorder—with Focus on the Cingulate Gyrus”: [https://substack.com/home/post/p-210137336]. Your PV-interneuron findings may provide the kind of cellular and temporal mechanism needed to connect local cortical dysfunction with these larger disturbances of network coordination and conscious experience.
In my opinion, schizophrenia is a disorder caused by a disruption of the brain’s predictive system.
You suggest that schizophrenia may involve impaired regulation and synchronization of cortical networks, rather than simply an excess or deficiency of a single neurotransmitter. This is probably true, although it does not by itself establish a privileged role for PV interneurons. The similarity between the symptoms of schizophrenia and transient ketamine-induced psychosis should perhaps be considered within the broader context of the brain’s adaptive and regulatory processes.
The analogy with Parkinson’s disease is heuristically useful, but limited. Schizophrenia is far more heterogeneous, encompassing different clinical courses, symptom profiles, and biological mechanisms.
Nevertheless, I would like to return to my original thesis: schizophrenia may be a disorder of hierarchical predictive inference, particularly of the way precision is assigned to predictions and prediction errors.
Several key components are involved in this chain of dependencies. Your observations concerning ketamine immediately suggest a computational metaphor in which dopamine and cortisol form part of a coupled system of reinforcement and correction, broadly resembling reinforcement-learning architectures. Dopamine may provide a rapid, directional prediction-error signal, while cortisol and the stress system may regulate the broader learning context: sensitivity to risk, the weighting of negative outcomes, and the rate at which beliefs are updated. Dysregulation of this coupling could lead to pathological assignments of precision and salience, forming a possible computational core of psychosis.
A possible progression would therefore be:
neural noise ⟶ incorrectly weighted prediction errors ⟶ attribution of significance to random events
If neutral stimuli repeatedly generate the signal “this is important,” the brain must infer a cause. A delusion may then emerge as an attempt to organize chaotic prediction errors, rather than as an arbitrary belief. In this sense, the delusion initially arises from excessive instability but later becomes exceptionally resistant to correction.
Hallucinations can similarly be interpreted as situations in which internally generated predictions are given too much weight relative to sensory evidence. A disturbance in predicting the consequences of one’s own actions, associated with efference copy or corollary discharge, may also produce the impression that one’s thoughts or actions originate from an external agent.
From this perspective, the PV-cell hypothesis does not compete with the predictive-processing hypothesis. It may describe one of its biological substrates. PV interneurons regulate gamma synchronization and the millisecond-scale temporal windows within which cortical signals reinforce, coordinate, or suppress one another. Their dysfunction need not simply increase overall brain activity. More importantly, it may impair the signal-to-noise ratio, temporal precision, and the encoding of informational reliability.
We should therefore consider a broader model in which schizophrenia is a heterogeneous disorder whose common computational core may be a dysregulation of hierarchical predictive processing: the miscoding of the precision, salience, and authorship of signals.
This framework explains more than either the dopamine hypothesis or PV dysfunction alone. Dopamine, NMDA receptors, GABAergic inhibition, gamma oscillations, stress regulation, and disrupted connectivity can then be understood as different levels of the same pathological process: biochemical, cellular, circuit-level, dynamic, and computational.
Thanks for your comments.
The schizophrenia story that is emerging is complex and fascinating. Unraveling the interplay between neural circuits and molecular mechanisms in the in the PFC is just beginning.
David Lewis has done the best research on the glutamate hypothesis. Ketamine is an NMDA receptor blocker, which triggers the downregulation of GAD67 in PV basket cells and somatostatin interneurons. The story is quite complex with interleukin 6 as an intermediate between them.
Dienel, S. J., Fish, K. N., & Lewis, D. A. (2023). The nature of prefrontal cortical GABA neuron alterations in schizophrenia: markedly lower somatostatin and parvalbumin gene expression without missing neurons. American Journal of Psychiatry, 180(7), 495-507.
Terry
Dr. Sejnowski, this is an excellent and important contribution. I particularly appreciate the move away from viewing schizophrenia primarily through the dopamine hypothesis and toward a model involving inhibitory interneurons, gamma coordination, and network timing. Your findings concerning GAD67 reduction and parvalbumin basket-cell dysfunction provide a plausible biological mechanism by which cortical activity could become both hyperactive and poorly synchronized.
In my own theoretical work on schizophrenia as a disorder of conscious moment formation, I have suspected that positive symptoms may be associated especially with posterior cingulate cortex and default-mode network dysfunction. The PCC contributes contextual integration, self-referential processing, and construction of an internal model of reality; impaired inhibitory timing in these networks might therefore allow internally generated representations to compete abnormally with sensory reality. I have also wondered whether the brain may compensate for this posterior instability by suppressing broader network propagation, thereby reducing hallucinations at the cost of producing secondary negative symptoms such as avolition, flattened affect, and diminished behavioral initiation. Have you encountered physiological or imaging evidence supporting either of these anterior-posterior or compensatory trends?
I discuss this possibility more fully in my Substack article, “Schizophrenia as a Network Disorder—with Focus on the Cingulate Gyrus”: [https://substack.com/home/post/p-210137336]. Your PV-interneuron findings may provide the kind of cellular and temporal mechanism needed to connect local cortical dysfunction with these larger disturbances of network coordination and conscious experience.
In my opinion, schizophrenia is a disorder caused by a disruption of the brain’s predictive system.
You suggest that schizophrenia may involve impaired regulation and synchronization of cortical networks, rather than simply an excess or deficiency of a single neurotransmitter. This is probably true, although it does not by itself establish a privileged role for PV interneurons. The similarity between the symptoms of schizophrenia and transient ketamine-induced psychosis should perhaps be considered within the broader context of the brain’s adaptive and regulatory processes.
The analogy with Parkinson’s disease is heuristically useful, but limited. Schizophrenia is far more heterogeneous, encompassing different clinical courses, symptom profiles, and biological mechanisms.
Nevertheless, I would like to return to my original thesis: schizophrenia may be a disorder of hierarchical predictive inference, particularly of the way precision is assigned to predictions and prediction errors.
Several key components are involved in this chain of dependencies. Your observations concerning ketamine immediately suggest a computational metaphor in which dopamine and cortisol form part of a coupled system of reinforcement and correction, broadly resembling reinforcement-learning architectures. Dopamine may provide a rapid, directional prediction-error signal, while cortisol and the stress system may regulate the broader learning context: sensitivity to risk, the weighting of negative outcomes, and the rate at which beliefs are updated. Dysregulation of this coupling could lead to pathological assignments of precision and salience, forming a possible computational core of psychosis.
A possible progression would therefore be:
neural noise ⟶ incorrectly weighted prediction errors ⟶ attribution of significance to random events
If neutral stimuli repeatedly generate the signal “this is important,” the brain must infer a cause. A delusion may then emerge as an attempt to organize chaotic prediction errors, rather than as an arbitrary belief. In this sense, the delusion initially arises from excessive instability but later becomes exceptionally resistant to correction.
Hallucinations can similarly be interpreted as situations in which internally generated predictions are given too much weight relative to sensory evidence. A disturbance in predicting the consequences of one’s own actions, associated with efference copy or corollary discharge, may also produce the impression that one’s thoughts or actions originate from an external agent.
From this perspective, the PV-cell hypothesis does not compete with the predictive-processing hypothesis. It may describe one of its biological substrates. PV interneurons regulate gamma synchronization and the millisecond-scale temporal windows within which cortical signals reinforce, coordinate, or suppress one another. Their dysfunction need not simply increase overall brain activity. More importantly, it may impair the signal-to-noise ratio, temporal precision, and the encoding of informational reliability.
We should therefore consider a broader model in which schizophrenia is a heterogeneous disorder whose common computational core may be a dysregulation of hierarchical predictive processing: the miscoding of the precision, salience, and authorship of signals.
This framework explains more than either the dopamine hypothesis or PV dysfunction alone. Dopamine, NMDA receptors, GABAergic inhibition, gamma oscillations, stress regulation, and disrupted connectivity can then be understood as different levels of the same pathological process: biochemical, cellular, circuit-level, dynamic, and computational.
Update: https://jacekhoffman.substack.com/p/schizophrenia-as-a-disorder-of-the