Phonetic Learning and Cultural Transmission in Accent Acquisition and Sound Symbolism

Human speech learning presents an apparent paradox. Infants begin with broad sensitivity to speech contrasts but become rapidly attuned to the distributions of their linguistic environment; adults remain capable of adapting to unfamiliar talkers and accents, although that adaptation is constrained by prior experience. Sound-symbolic phenomena such as the bouba–kiki effect reveal a complementary pattern: cross-modal regularities recur across languages and, in some cases, outside human linguistic experience. Neural language and speech models complicate both observations because they can exhibit human-like phonological or sound-symbolic behavior without sharing human developmental, sensorimotor, or evolutionary histories. This perspective argues that these findings are best unified not by a literal “universal grammar of accent,” but by probabilistic phonetic priors: broadly shared auditory, perceptual, articulatory, cross-modal, and learning constraints that can be culturally canalized into community-specific phonetic systems. The proposed Phonetic Inheritance Framework distinguishes three causal sources of phonetic structure, architectural/inductive bias, perceptual/physical bias, and cultural/data inheritance, and introduces phonetic cultural fossils as statistical traces of human perceptual and articulatory pressures preserved in linguistic data and recoverable by artificial learners. The central inferential principle is that similar human and model outputs do not establish a shared causal origin. The framework treats accent acquisition and sound symbolism as complementary probes and derives testable predictions involving decontaminated training, modality ablation, cross-linguistic transfer, synthetic input, and developmental-trajectory comparisons.