Simulating usage-based sound change: How perceptual confusion shapes vowel dynamics

Abstract

In usage-based linguistics, models of sound change (e.g., exemplar theory) frequently rely on a perfect strong categorical perception (CP) mechanism (i.e., invariably selecting the highest-probability candidate). This approach overlooks a crucial feature of human perception: perceptual confusion arising from fuzzy boundaries between acoustically similar sounds. To test whether such a cognitively plausible weak CP mechanism generates significantly different results than the traditional strong CP baseline, I integrated a probabilistic decision rule into agent-based simulations. Manipulating population size and acoustic noise, results revealed that weak CP yielded a significantly more robust system. Specifically, it maintained higher stability in high-noise environments, achieved greater consensus in large populations, and substantially enhanced communication efficiency and cognitive economy. This study concludes that perceptual confusion must no longer be treated as mere noise, but rather as a key adaptive mechanism. Consequently, this micro-level instability proves essential for driving system-wide robustness across the linguistic community.