Task Contrasts Dissociate Across Neural-Dynamics Estimators

This article presents an exploratory within-subject reanalysis of the OpenNeuro ds004511 EEG dataset under Noetic Diffusion Theory (NDT). The analysis asks whether Rest, cognitive control (CC), and gambling (GG) task contrasts agree across explicitly non-interchangeable summaries of a subject-anchored neural chart. The frozen export contains 134 recordings from 45 participants: 45 Rest, 45 CC, and 44 GG recordings. The measurement contract uses the canonical MNPS [m,d,e] chart and its stratified 9D definition with subject-anchored coordinates. The article reports chart occupancy, Meta-Noetic Jacobian (MNJ) quantities, four L1 local law measures, chart-Jacobian spectral abscissa, identified lag-1 one-step L2 and L3 quantities, and certificate-derived L4 reachability amplitudes. These are treated as distinct estimators rather than interchangeable labels. GG minus Rest increased chart-Jacobian alpha (median difference 0.01248, paired dz = 1.42, n = 44, BH q = 5.61e-10). The direction persisted after re-aggregating every recording over the same first 400 seconds. The independently identified one-step L3 generator-proxy spectral abscissa did not show a positive counterpart (dz = -0.328, n = 32, q = 0.178), while identified L2 and L3 rotation summaries decreased. The complete L1 family is reported without significance-based row selection; local chart-drift norm effects were large and remained positive in the deterministic 400-second sensitivity, whereas several diffusion effects attenuated. A separate exploratory estimator-by-task interaction analysis on 121 recordings with common production support found positive GG-Rest and GG-CC interactions after estimator-specific, outcome-blind robust standardization. This estimand is normalization-dependent and does not compare raw endpoint units. A frozen subject-level concordance diagnostic found weak or negative Spearman associations between standardized chart and L3 task contrasts (GG-Rest rho = -0.383; 12/32 same-sign pairs). Reachability amplitude contrasts remain duration-limited because task duration ranges do not overlap and L4 is exported only as recording-level scalars. The degenerate effective-dimension readout is retained as QC and is not interpreted as a dimension result. The findings establish an estimator-specific dissociation in one frozen export, not a biological mechanism, causal task effect, clinical biomarker, consciousness measure, or confirmatory result. Public supplements document the measurement contract, complete endpoint families, missingness, robustness analyses, and reproducibility boundaries.