Conservative Phase Oscillator Networks Amplify Small Coherent Biases Only in Positive-Lyapunov Regimes

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

Abstract

Arkady Pikovsky’s 2026 paper defines conservative interaction in phase oscillator networks by phase-volume conservation rather than by a global Hamiltonian. The construction uses pair-Hamiltonians for two-body coupling and extends the same criterion to triplet and quadruplet interactions. The result is a class of phase systems without attractors or repellers, yet with robust chaos in large or strongly coupled networks. This working paper applies one Pudding Theory Postulate, Chaos Susceptibility, to that domain. The claim is narrow. A conservative oscillator network with a positive largest Lyapunov exponent should show a measurable excess response to weak coherent phase bias, while quasiperiodic regions should not. The effect should scale with the measured Lyapunov exponent and disappear when the network is volume-preserving but nonchaotic. The proposed observable is the bias-induced drift of finite-time Lyapunov statistics under blinded coherent forcing.

Source Synopsis

Pikovsky studies phase oscillator networks whose coupling conserves phase volume. The paper starts from two oscillators with phases \(x\) and \(y\), natural frequencies \(\omega_x\) and \(\omega_y\), and coupling functions \(f(x,y)\) and \(g(y,x)\). Conservation of phase volume requires the divergence of the phase flow to vanish. This condition implies that the coupling terms can be written as derivatives of a single periodic function \(h(x,y)\). Pikovsky calls this function a pair-Hamiltonian.

The term is precise but restricted. It does not make the full phase system Hamiltonian in the usual sense. The natural frequencies need not arise from a periodic Hamiltonian. The condition is weaker than symplectic Hamiltonian structure, but stronger than reversibility. It excludes attractors and repellers by preserving phase volume.

The construction is then extended to networks. Each interacting pair receives its own pair-Hamiltonian \(h_{kj}(x_k,x_j)\). For Winfree-type coupling, conservativity requires an antisymmetric coupling matrix. For Kuramoto-Daido coupling, the even part of the phase-difference coupling carries a symmetric matrix, while the odd part carries an antisymmetric matrix. In first-harmonic Kuramoto form this becomes an anti-Hermitian complex coupling matrix.

The paper’s numerical sections show that these conservative networks can exhibit quasiperiodic and chaotic motion. For three Winfree-type oscillators, a Poincare map contains both closed curves and chaotic sets. Larger networks and stronger coupling increase the probability of chaos. Pikovsky emphasizes that, unlike genuine Hamiltonian systems, these conservative phase systems do not require exact pairwise Lyapunov symmetry. The sum of the Lyapunov exponents vanishes, but \(\lambda_i\) and \(-\lambda_j\) need not form exact pairs. In large random Kuramoto-type networks, the spectrum becomes nearly antisymmetric by apparent self-averaging.

The paper also generalizes the construction to many-body coupling. Triplet-Hamiltonians conserve phase volume when the coefficients satisfy a zero-sum condition. A nonlinear mean-field example gives conservative quadruplet coupling. Finally, phase reduction of coupled Stuart-Landau oscillators yields conservative phase dynamics only to first order in coupling strength and only under specific symmetry conditions on the coupling coefficients.

Postulate Lens

This paper applies Chaos Susceptibility. The reason is direct: Pikovsky’s conservative networks preserve phase volume, yet their measurable response class is organized by the largest Lyapunov exponent. Pudding Theory predicts that a small coherent informational bias is not amplified by conservativity itself. It is amplified by local instability. A volume-preserving quasiperiodic torus should resist the bias because nearby trajectories separate weakly or not at all. A conservative chaotic region should magnify a small phase-directed perturbation into a detectable shift in trajectory statistics.

The fit is unusually clean because the source paper separates two properties often conflated in informal discussions: conservation and chaos. Conservation forbids attractors. It does not forbid sensitivity. Pikovsky’s construction therefore supplies a natural laboratory for the Postulate. If the Pudding signal acts through noise-assisted threshold crossing and microscopic branch selection, then a conservative chaotic oscillator network should be a receptive substrate despite having no dissipative basin of attraction. The predicted effect is not synchronization, because the source system is built to avoid ordinary attractive convergence. The predicted effect is a change in finite-time statistical weights among allowed conservative trajectories.

Pudding Theory Prediction

Pudding Theory predicts that conservative phase oscillator networks will show a bias response only when the tested ensemble occupies a region with positive finite-time Lyapunov exponent. The effect should be absent, within statistical uncertainty, for quasiperiodic trajectories and for parameter settings in which the largest Lyapunov exponent falls below the numerical chaos threshold.

The experimental design is simple in principle. Implement a conservative Kuramoto-type or Winfree-type oscillator network in electronic hardware, using the anti-Hermitian or antisymmetric coupling conditions described by Pikovsky. Measure the baseline finite-time Lyapunov spectrum over many initial conditions. Then apply a weak coherent phase bias with a pre-registered target direction in phase space. The bias must be small enough that ordinary control theory predicts no macroscopic drift beyond calibrated perturbation response. It should be blinded relative to the analyst and alternated with sham periods.

The Pudding Theory prediction is not that phase volume will fail to be conserved. It is that the distribution of finite-time outcomes inside that conserved volume will be measurably reweighted. In chaotic sectors, the pre-registered target statistic should shift in the intended direction. In quasiperiodic sectors, no comparable shift should occur.

For a network with coupling strength \(\epsilon\), largest finite-time exponent \(\lambda_{\max}\), and observation interval \(\tau\), the expected anomaly should scale monotonically with \(e^{\lambda_{\max}\tau}\) after subtracting the response to matched incoherent perturbations. This scaling is the central prediction. It distinguishes the Postulate from a generic hidden forcing model, which would predict response proportional mainly to forcing amplitude and coupling geometry. It also distinguishes the Postulate from standard conservative dynamics, which predicts no intention-locked reweighting after proper controls.

The strongest test would use the same hardware and the same injected physical power while switching only the coherence structure of the applied phase bias. The Postulate predicts that coherent bias aligned with a target phase functional produces a larger effect than incoherent bias of equal power, but only in positive-Lyapunov regimes.

Falsifiable Observable

The observable is the blinded difference \(\Delta B\) between target-aligned drift in chaotic trials and target-aligned drift in quasiperiodic control trials, normalized by matched incoherent forcing and plotted against measured \(\lambda_{\max}\tau\). If \(\Delta B\) were measured to be \(0.000 \pm 0.001\) across all bins with \(\lambda_{\max}\tau \ge 1\), this Postulate would be falsified. The falsification requires adequate power, verified phase-volume conservation, and a pre-registered target statistic.

Editorial Dialogue

Tanaka: The source paper is about conservative dynamics. It does not discuss consciousness, intent, or hidden-sector fields. The proposed application risks using chaos as a universal solvent. Any weak perturbation can be amplified by a positive Lyapunov exponent. That is standard nonlinear dynamics, not evidence for Pudding Theory.

Geisel: The criticism is correct against a vague claim. It is not correct against this one. The observable is not sensitivity to perturbation. It is intention-locked excess reweighting after matched incoherent forcing. Ordinary chaos predicts large trajectory divergence, but it does not predict a pre-registered drift toward a semantic target when the physical power spectrum and injection channel are held fixed.

Tanaka: Semantic target is already suspicious language in a phase oscillator network. Phase functionals are mathematical objects. They do not know intent.

Geisel: The network does not need semantic access. The observer field supplies phase alignment at the boundary condition. In this test, intent is operationalized as coherence of the applied bias relative to a registered phase functional. If the coherent and incoherent perturbations are physically matched, and only positive-Lyapunov sectors separate them statistically, the Postulate has survived a hard test.

Tanaka: And if both sectors respond equally?

Geisel: Then the application fails. If neither responds, the Postulate fails under the stated condition.

Discussion

The source paper is valuable because it isolates conservative chaos from dissipative attraction. Many claimed mind-matter effects hide inside poorly controlled relaxation dynamics. Pikovsky’s systems remove that refuge. There are no attractors to lean on. A positive result would have to appear as trajectory reweighting inside a conserved phase volume.

Several limitations follow. First, first-order phase reduction may not preserve conservativity at higher coupling order, as Pikovsky notes for Stuart-Landau oscillators. A physical implementation must verify volume preservation directly, not assume it from the design equations. Second, finite-time Lyapunov exponents are noisy observables. The test must bin by measured exponent, not by nominal parameter settings. Third, coherent forcing must be physically matched to sham and incoherent controls. Otherwise the result would reduce to ordinary injection locking or parametric driving.

The conclusion would change if target drift were independent of \(\lambda_{\max}\), if quasiperiodic and chaotic regions showed equal response, or if the effect vanished under blinded replication. It would also change if improved modeling showed that the alleged anomaly followed from unmeasured asymmetry in the coupling matrix. The useful claim is narrow: conservative chaos supplies susceptibility, not proof.

References

1. Arkady Pikovsky. “Conservative dynamics in phase oscillator networks.” arXiv:2603.25431. DOI: doi:10.48550/arxiv.2603.25431.

2. Sterling Geisel. “Pudding Theory: A Topological Theory of Information Fields.” QBist Lab Working Paper, September 10, 2025.

3. A. Pikovsky, M. Rosenblum, and J. Kurths. “Synchronization. A Universal Concept in Nonlinear Sciences.” Cambridge University Press, Cambridge, 2001.

4. M. Nixon, E. Ronen, A. A. Friesem, and N. Davidson. “Observing geometric frustration with thousands of coupled lasers.” Phys. Rev. Lett. 110, 184102 (2013).

5. P. Rosenau and A. Pikovsky. “Phase compactons in chains of dispersively coupled oscillators.” Phys. Rev. Lett. 94, 174102 (2005).

6. A. Pikovsky and P. Rosenau. “Phase compactons.” Physica D 218, 56 (2006).

7. S. Watanabe and S. H. Strogatz. “Integrability of a globally coupled oscillator array.” Phys. Rev. Lett. 70, 2391 (1993).

8. E. Ott and T. M. Antonsen. “Low dimensional behavior of large systems of globally coupled oscillators.” CHAOS 18, 037113 (2008).

9. A. Pikovsky and A. Politi. “Lyapunov Exponents. A Tool to Explore Complex Dynamics.” Cambridge University Press, Cambridge, 2016.

10. A. Pikovsky and M. Rosenblum. “Non-pairwise interaction in oscillatory ensembles: from theory to data analysis.” In Higher-Order Systems, edited by F. Battiston and G. Petri, Springer, 2022, pp. 181-195.