What must a theory of perturbational complexity explain? Nine preregistration-hardened constraints, six dead hypotheses, and a minimal visibility account
Year: 2026
Type: preprint
Abstract: Across five preprints, this program measured how perturbational complexity behaves once its estimator is debiased, what quantity it tracks (the reproducible dimensionality of the evoked response, R-dim), and how that quantity depends on dynamics, trial count, species and arousal state - always under preregistrations sealed before data contact, with failures published. This paper consolidates the empirical residue into nine constraints that any theory of perturbational complexity must satisfy, reports the autopsies of six of the author's own hypotheses that died against them (including a formal theory killed in two sealed prediction cycles and buried unpublished), and offers the minimal account known to survive: reproducible components with rapidly decaying amplitudes become visible one by one as trial averaging lowers an effective noise floor. With an approximately exponential amplitude spectrum, three lines of algebra reproduce five constraints at once: logarithmic growth of R-dim with trial count, the absence of a detectable human ceiling at protocol-scale trial counts, finite system-specific ceilings in simulated networks, level-but-not-growth scaling with electrode coverage, and - via uniform attenuation - the parallel downward displacement of the growth curve under anesthesia (level, not slope). What the account does not explain is listed with equal clarity. It also makes one quantitative, falsifiable prediction (P1): that the displacement between two states' growth curves follows from the evoked-amplitude ratio alone. Sealed publicly and run once, it PASSED at the field level (ordering rho = +0.41, one-sided p = 0.038, n = 20; median observed/predicted ratio 1.00) and then FAILED its cross-level replication in the spiking populations of the same brains (sealed separately before the spiking gain was computed; ordering inverted to rho = -0.29, and observed displacements exceeded predictions by a median factor of 3.0): anesthesia does more to neuronal reproducible dimensionality than uniform amplitude gain can account for. The account therefore survives as an economical summary of the trial-scaling constraints and dies as a mechanistic claim at the level where the mechanism lives - the refutation of the author's own model is reported here rather than left for others. Two further sealed protocols chase the structural change itself. Anesthesia rotates the population response subspace to near-orthogonality (median between-state overlap 0.082 vs a within-state control of 0.532; lower in 15 of 15 animals, p = 3.1e-5): anesthesia does not attenuate a fixed population response so much as replace it. Yet the extent of rotation does not scale the excess displacement either (rho = -0.09), killing rotation-as-explanation in the same run that established the rotation. The last pre-declared candidate, per-component decoherence, is reported as untested: the estimator of trial-level coherence at matched amplitude failed its pre-registered known-truth validation bench in three versions (criteria fixed before each run, all archived), and under the program's rules an uncertified instrument runs no confirmatory test. No new data; all constraints reference published analyses with public preregistrations, per-subject tables, code archives and mechanical verifiers.
Author: Galindez, Nicolás Federico
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7204705948
Cites:
Cited by:
Related to: