Associating the origin and spread of sound change using agent-based modelling applied to /s/-retraction in English

Abstract

The study explored whether an asymmetric phonetic overlap between speech sounds could be turned into sound change through propagation around a community of speakers. The focus was on the change of /s/ to /ʃ/ which is known to be more likely than a change in the other direction both synchronically and diachronically. An agent-based model was used to test the prediction that communication between agents would advance /s/-retraction in /str/ clusters (e.g. string). There was one agent per speaker and the probabilistic mapping between words, phonological classes, and speech signals could be updated during communication depending on whether an agent listener absorbed an incoming speech signal from an agent talker into memory. Followinginteraction, sibilants in /str/ clusters were less likely to share a phonological class with prevocalic /s/ and were acoustically closer to /ʃ/. The findings lend support to the idea that sound change is the outcome of a fortuitous combination of the relative size and orientation of phonetic distributions, their association to phonological classes, and how these types of information vary between speakers that happen to interact with each other.