Harmonic Brain Flows Are Material Memory Traces in Counterfactual Network Dynamics

Authors: Sterling Geisel, QBist Lab; Dr. Hideo Tanaka

Abstract

Chung, Maccotta, and Struck model brain causality as directed edge flow on a simplicial scaffold and identify the harmonic component as the persistent backbone that survives dissipative decay. Pudding Theory reads this harmonic component as a physical memory trace, not only as a mathematical invariant. Under the Material Memory Postulate, repeated neural signaling leaves conserved flow structure in matter. The harmonic subspace is therefore the observable carrier of accumulated signal history. Temporal lobe epilepsy appears, in this reading, as over-conserved limbic memory, where recurrent temporal and subcortical traces dominate the causal field. Surgical counterfactuals test whether the trace can be rerouted without erasing the global memory topology. The source paper treats harmonic flow as a zero-energy network component. Pudding Theory treats it as retained informational bias. If the post-intervention harmonic projection of repeatedly stimulated temporal-limbic pathways were measured to be statistically independent of prior recurrence history, this Postulate would be falsified.

Source Synopsis

Chung, Maccotta, and Struck introduce a counterfactual framework for brain network dynamics. Their target is a limitation of standard causal neuroimaging models. Granger causality, structural equation modeling, and dynamic causal modeling can infer directed association, but they do not directly answer what the causal organization would become after a pathway is disrupted, amplified, or therapeutically modulated.

The authors represent directed functional interactions as time-varying edge flows estimated from resting-state fMRI. AAL parcels supply 116 anatomical nodes. Time-lagged Pearson correlations over sliding windows supply directed edges. Because raw average directed correlations are weak across subjects and time, the authors build a sparse higher-order spatial scaffold using 2-simplices. Edges below a threshold are removed. Triangles are retained only when their edges pass the threshold, producing a directed complex that is much smaller than the full Rips complex.

The mathematical core is the 1-Hodge Laplacian. It decomposes edge flow into components associated with divergence, cyclic circulation, and harmonic persistence. The Dirichlet energy of a flow is defined through this Laplacian. Gradient descent under the energy functional dissipates all positive-eigenvalue modes. The remaining zero-eigenvalue modes form the harmonic projection. The authors describe this object as the stable, recurrent, non-dissipative backbone of brain communication.

Counterfactual operators then act on flows by scaling, deleting, or modulating selected connections. A large energy increase indicates fragility. Near-zero change indicates compensation through energetically equivalent routes. The authors apply this to temporal lobe epilepsy. They simulate pathological recurrence by amplifying temporal-limbic feedback by 30 percent. They simulate left anterior temporal lobectomy by attenuating left temporal-limbic and hippocampal-amygdalar flows by 70 percent. In healthy controls, lobectomy leaves dominant harmonic flows largely intact. In the disease model, recurrence shifts dominance toward olfactory, amygdalar, rectal, vermian, and subcortical loops.

Postulate Lens

The applied Postulate is Material Memory. Matter retains the trace of repeated signals, and the trace biases future probability.

The source system already has this structure. Recurrent neural signaling is not discarded as transient activity. It survives projection into the harmonic subspace. In the authors' terms, positive-energy modes decay while zero-energy modes remain. In Pudding Theory terms, repeated causal traffic has written a durable bias into the physical network. The harmonic flow is the measurable residue of that writing.

This Postulate fits because the source paper studies persistence under counterfactual perturbation. Its central observable is not momentary activation but the retained topology of recurrent communication.

Pudding Theory Reading

Pudding Theory reads the Hodge harmonic component as materialized signal history. The brain network is not only passing information through edges. It is storing prior traffic in the topology of allowable circulation. The harmonic projection is the place where repeated neural signals have become a constraint on future dynamics.

This changes the interpretation of the source model. In the source framing, the harmonic flow is the non-dissipative component of an edge-flow decomposition. It is a zero-energy mathematical backbone. Pudding Theory accepts the decomposition but assigns physical content to the invariant. A zero Dirichlet-energy flow is not empty of cause. It is saturated with past cause. It is a retained signal configuration whose energy cost has already been paid by prior organization of the tissue.

The important object is therefore not connectivity strength alone. It is recurrence that survives dissipation. A pathway that repeatedly carries temporally structured activity can become a probability well for later causal traffic. Future flows do not choose among equivalent pathways freely. They are biased toward routes already stabilized by prior signal history. The source paper treats the counterfactual operator as an external manipulation of a given network. Pudding Theory reads the response to that operator as a measurement of how deeply the network has remembered its previous use.

Temporal lobe epilepsy then acquires a sharper meaning. The disease model does not merely amplify temporal-limbic loops. It exposes an over-retained memory trace. Amygdalar, olfactory, rectal, vermian, and subcortical pathways become dominant because recurrent pathological signaling has made them the preferred harmonic routes. The seizure-prone network is not just hyperconnected. It is topologically biased toward recurrence. Its past episodes have altered the future probability landscape of communication.

The healthy lobectomy counterfactual is equally important. When left temporal-limbic pathways are attenuated and global harmonic organization remains stable, the network is not indifferent to damage. It has distributed memory. Contralateral and midline routes carry enough retained signal structure to preserve the global pattern. Compensation is not only rerouting after injury. It is the activation of stored alternative traces already present in the material network.

The structural claim is direct: the harmonic basis is constrained by repeated use. In a brain with recurrent temporal-limbic pathology, the harmonic projection after perturbation should remain biased toward the history of pathological recurrence, even when instantaneous correlation strength is controlled. What the source treats as the invariant backbone, Pudding Theory treats as the material memory register.

Falsifiable Observable

The distinguishing observable is the dependence of post-intervention harmonic projection on prior recurrence history after controlling for baseline edge weights and instantaneous correlation. In longitudinal TLE data, pathways repeatedly recruited during seizures should occupy the harmonic projection more strongly after virtual or real intervention than equally weighted pathways without such recurrence history. If the post-intervention harmonic projection of repeatedly stimulated temporal-limbic pathways were measured to be statistically independent of prior recurrence history, this Postulate would be falsified.

Editorial Dialogue

Tanaka: The reading risks reifying a decomposition. Hodge theory supplies a useful basis, but a harmonic component need not be a physical memory. It may be a coordinate artifact of the chosen simplicial scaffold, the AAL parcellation, and the thresholding rule. If the scaffold changes, the harmonic subspace changes. Why call that material memory?

Sterling: Because the claim is not that every computed harmonic vector is ontologically final. The claim is that persistence under dissipation is the correct place to look for retained signal. The source model already separates transient flow from recurrent flow. Pudding Theory says the recurrent component is not only descriptive. It is the substrate through which prior signaling biases future causal organization.

Tanaka: But prior seizure activity is not directly included in the paper’s healthy HCP dataset. The TLE model is simulated by amplification.

Sterling: That is why the falsifier is longitudinal. The source framework gives the operator and the observable. The Pudding reading adds a constraint: true pathological recurrence should leave history-dependent harmonic occupation. A pure energy-perturbation account can treat two equally weighted pathways as exchangeable. Material Memory says they are not exchangeable if one has carried repeated pathological signal.

Discussion

The reading buys a physical interpretation of compensation. In the source paper, compensation is energetic and topological. The network preserves zero-energy harmonic organization by shifting flow through equivalent routes. Pudding Theory adds that equivalence is constrained by retained signal history. Some routes are available because matter has already stored them as low-resistance causal circulation.

This matters for epilepsy. A seizure network may look abnormal not only because current synchronization is excessive, but because past synchronization has hardened future recurrence. Treatment then becomes a question of memory topology. A successful intervention should not only reduce pathological edge strength. It should weaken the harmonic occupation of recurrent temporal-limbic traces and increase occupation of compensatory contralateral or midline traces.

The limitation is measurement. Resting-state fMRI estimates are indirect, slow, and scaffold-dependent. The claim requires longitudinal data, seizure-history annotations, and robustness across parcellations and threshold choices. If recurrence history fails to predict harmonic persistence once edge weights are controlled, the Pudding reading fails. If it succeeds, harmonic flow becomes more than a network invariant. It becomes the measurable trace of prior signal written into brain matter.

References

1. Moo K. Chung, Luigi Maccotta, and Aaron F. Struck. “Counterfactual Analysis of Brain Network Dynamics.” arXiv:2603.29843, 2026. DOI: doi:10.48550/arxiv.2603.29843.

2. S. Ochs. “Pudding Theory: A Topological Theory of Information Fields.” QBist Lab Working Paper, 2026.

3. D. V. Anand and M. K. Chung. “Hodge-Laplacian of Brain Networks.” IEEE Transactions on Medical Imaging 42, 1563-1473, 2023.

4. F. Bartolomei, S. Lagarde, F. Wendling, A. McGonigal, V. Jirsa, M. Guye, and C. Benar. “Defining Epileptogenic Networks: Contribution of SEEG and Signal Analysis.” Epilepsia 58, 1131-1147, 2017.

5. M. K. Chung, C. G. Ramos, F. B. De Paiva, J. Mathis, V. Prabhakaran, V. A. Nair, M. E. Meyerand, B. P. Hermann, J. R. Binder, and A. F. Struck. “Unified Topological Inference for Brain Networks in Temporal Lobe Epilepsy Using the Wasserstein Distance.” NeuroImage 284, 120436, 2023.

6. K. J. Friston, A. M. Bastos, A. Oswal, B. Van Wijk, C. Richter, and V. Litvak. “Granger Causality Revisited.” NeuroImage 101, 796-808, 2014.

7. H. F. J. Gonzalez, S. Chakravorti, S. E. Goodale, K. Gupta, D. O. Claassen, B. Dawant, V. L. Morgan, and D. J. Englot. “Thalamic Arousal Network Disturbances in Temporal Lobe Epilepsy and Improvement After Surgery.” Journal of Neurology, Neurosurgery & Psychiatry 90, 1109-1116, 2019.

8. D. C. Van Essen, K. Ugurbil, E. Auerbach, D. Barch, T. E. J. Behrens, R. Bucholz, A. Chang, L. Chen, M. Corbetta, and S. W. Curtiss. “The Human Connectome Project: A Data Acquisition Perspective.” NeuroImage 62, 2222-2231, 2012.