Spatial-Cognitive Schema-Based Interactive Language Training: An In-Silico Computational Simulation of Voice-Onset Latency and Novel-Sentence Generalization

Abstract

# Spatial-Cognitive Schema-Based Interactive Language Training: An In-Silico Computational Simulation of Voice-Onset Latency and Novel-Sentence Generalization **Author:** Damian (Independent Researcher)**Date:** 2025**Type:** Preprint / Computational Simulation Study --- ## Abstract This study presents a pre-specified in-silico computational simulation examining the effects of spatial-cognitive schema-based interactive language training on voice-onset latency (VOL) and novel-sentence generalization accuracy. Sixty virtual cognitive agents (30 per condition) were parameterized using Gaussian latency distributions and binomial accuracy probabilities derived from cognitive load theory. Group A (3D Spatial Schema) was modeled with N(μ=420, σ=45) ms latency and p=0.91 accuracy; Group B (2D Translation/Rule) was modeled with N(μ=680, σ=75) ms latency and p=0.73 accuracy. Each agent completed 50 novel-sentence trials (total: 3,000 observations). A Monte Carlo robustness check comprising 10,000 independent simulation iterations confirmed 100% statistical significance across both endpoints. ## Key Results | Metric | Group A (3D Schema) | Group B (2D Translation) ||:---|:---:|:---:|| Mean Response Latency | 415.81 ms (SD=54.67) | 676.00 ms (SD=78.04) || Novel-Sentence Accuracy | 89.7% | 71.6% || Statistical Significance | p = 1.83 × 10⁻²⁰ | p = 1.03 × 10⁻³⁵ || Effect Size | Cohen's d = 3.86 | Odds Ratio = 3.44 | **Monte Carlo Robustness (10,000 iterations):**- Significance rate: 100.0% (both endpoints)- Mean latency difference: -260.16 ms (95% CI: [-291.09, -229.42] ms)- Mean accuracy difference: +18.00 %p (95% CI: [15.33, 20.73] %p) ## Ethics Statement This study does **not** involve human participants, human tissues, or human data. No institutional review board (IRB) approval was required. No external AI model APIs were invoked to generate cognitive response data. All simulated data are produced from mathematically specified probability distributions (Gaussian and Binomial) within a local computational environment. This manuscript explicitly disclaims any representation of human behavioral data. ## Keywords cognitive schema; voice-onset latency; novel-sentence generalization; in-silico simulation; Monte Carlo method; computational cognitive modeling; translation dependency ## Reproducibility All simulations used fixed random seeds (Seed = 2026 for the primary analysis; Seed = 8888 for the Monte Carlo check). The complete Python source code is provided as supplementary material in this Zenodo record. ## License Creative Commons Attribution 4.0 International (CC-BY 4.0)