Search for a command to run...
Nonlinear field-theoretic model of “meaning activation and conscious selection” on a semantic graph. A meaning manifold is constructed from fixed sentence embeddings (sentence-transformers/all-MiniLM-L6-v2, 384-d). Dynamics are modeled by a focusing nonlinear Schrödinger / Gross–Pitaevskii equation on a k-NN graph with an external potential derived from an empirical probability measure μ (frequency/salience). The selection time T* is defined operationally as the first hitting time of a peak-density threshold; a minimal observation model links it to reaction time as RT = T* + ε. The main testable prediction is a nonlinear acceleration of the right tail of RT distributions (vincentile regression) under high-arousal prime × ultra-short SOA conditions.Files: PDF (main text), DOCX (editable), README (structure and citation).AI assistance: AI language models (Anthropic Claude and xAI Grok) were used in a limited way for editorial support and organizational suggestions. The author takes full responsibility for the scientific content and any remaining errors. v4 (2026-03-07): This version has been substantially revised relative to earlier releases. The title has been updated to reflect the revised scope and framing. Main changes include updated figures, revised text structure, and refreshed supplementary materials. Please cite the version-specific DOI for this release. V5: Refined the theoretical framing and internal consistency of the SQM manuscript. The role of initial energy E[Φ0]E[\Phi_0]E[Φ0] is now stated explicitly as model intuition rather than as an analytically established or empirically validated predictor, with the current simulations showing only a weak and non-significant directional association with T∗T^*T∗. The phase assignment has been reformulated using a continuous atan2(x ⋅ u, x ⋅ c)\operatorname{atan2}(x\!\cdot\!u,\;x\!\cdot\!c)atan2(x⋅u,x⋅c) construction, and the dependence on the auxiliary quadrature direction uuu is now acknowledged explicitly. The SOA-proxy discussion has also been corrected to state that the present σ\sigmaσ-based mapping does not reproduce the intended SOA ordering, motivating a dedicated SOA parameter in future work. In addition, comparison with alternative RT models has been made more cautious by restricting claims to the present parameterisation and to qualitative shape differences in upper-tail amplification.