Linking Cognitive and Social Aspects of Sound Change Using Agent‐Based Modeling

The paper defines the core components of an interactive‐phonetic (IP) sound change model. The starting point for the IP‐model is that a phonological category is often skewed phonetically in a certain direction by the production and perception of speech. A prediction of the model is that sound change is likely to come about as a result of perceiving phonetic variants in the direction of the skew and at the probabilistic edge of the listener's phonological category. The results of agent‐based computational simulations applied to the sound change in progress, /u/‐fronting in Standard Southern British, were consistent with this hypothesis. The model was extended to sound changes involving splits and mergers by using the interaction between the agents to drive the phonological reclassification of perceived speech signals. The simulations showed no evidence of any acoustic change when this extended model was applied to Australian English data in which /s/ has been shown to retract due to coarticulation in /str/ clusters. Some agents nevertheless varied in their phonological categorizations during interaction between /str/ and /ʃtr/: This vacillation may represent the potential for sound change to occur. The general conclusion is that many types of sound change are the outcome of how phonetic distributions are oriented with respect to each other, their association to phonological classes, and how these types of information vary between speakers that happen to interact with each other. Using agent‐based modelling, Harrington, Kleber, Reubold, Schiel & Stevens (2018) develop a unified model of sound change based on cognitive processing of human speech and theories of how social factors constrain the spread of change throughout a community. They conclude that many types of change result from how biases in the phonetic distribution of phonological categories are transmitted via accommodation processes between individuals in interaction.