Sensitivity Analysis of Perceptual Stabilization under Historical Compression

Abstract

This paper examines parameter sensitivity in a toy computational realization of perceptual stabilization under historical compression. The analysis varies state coordinates, coupling coefficients, thresholds, and selector–dominance relations while preserving the declared computational equations and their status semantics. The study evaluates separate coordinates for route formation, stabilization level, perceptual-stabilization status, current-signal support, support discordance, selector–dominance alignment, and a restricted false-completion interpretation. Eight deterministic sensitivity families are combined with three episode/state Monte Carlo strata of 20,000 cases each and one 50,000-case parameter-family Monte Carlo study, all using declared ranges, seeds, denominators, and reporting rules. Across every Monte Carlo family, all four common-defined combinations of perceptual stabilization and signal support occur, demonstrating that the two statuses are not definitionally identified in the realization. In the multi-block isolation sweep, all 40,401 cases are support-discordant while only 30,301 satisfy the additional conditions for the narrower false-completion flag, establishing that support discordance and false completion are computationally non-identical in this model. The results characterize only the declared equations and sampling laws. They do not establish empirical prevalence, biological validity, universal parameter robustness, or probabilities for cognitive systems. Undefined states remain distinct from negative classifications throughout the analysis.