Phonetic accommodation and inhibition in a dynamic neural field model

TL;DR

The paper addresses how real-time input from another speaker shapes speech planning representations and contributes to accent change. It introduces a dual-layer dynamic neural field model that links online planning $u(x,t)$ with a slower memory field $u_{memory}(x,t)$, where excitatory/inhibitory interactions and a memory gating mechanism produce convergence during shadowing and, via delayed memory inhibition, possible divergence on subsequent trials. Empirical data from a pilot shadowing study show substantial individual variability with small, vowel-specific effects, which the authors use to motivate simulations. The simulations demonstrate that vowel-specific inhibitory memory dynamics can reproduce patterns of convergence followed by return to baseline or divergence, offering a mechanistic link between short-term accommodation and long-term sound change, and highlighting sociolinguistic pressures as modulators of memory inhibition. Overall, the work connects micro-level phonetic adaptation to macro-level accent dynamics through a quantitatively constrained, time-scale–aware neural field framework, with implications for how exposure and memory interact to shape speech over time.

Abstract

Short-term phonetic accommodation is a fundamental driver behind accent change, but how does real-time input from another speaker's voice shape the speech planning representations of an interlocutor? We advance a computational model of change in speech planning representations during phonetic accommodation, grounded in dynamic neural field equations for movement planning and memory dynamics. A dual-layer planning/memory field predicts that convergence to a model talker on one trial can trigger divergence on subsequent trials, due to a delayed inhibitory effect in the more slowly evolving memory field. The model's predictions are compared with empirical patterns of accommodation from an experimental pilot study. We show that observed empirical phenomena may correspond to variation in the magnitude of inhibitory memory dynamics, which could reflect resistance to accommodation due to phonological and/or sociolinguistic pressures. We discuss the implications of these results for the relations between short-term phonetic accommodation and sound change.

Figures (3)

• Figure 1: By-speaker difference in distance values for vowel and block. Values are from the Bayesian model and represent each speaker's random intercept coefficient added to the model's grand intercept. Negative values indicate convergence towards the model talker (blue shading); positive values indicate divergence (red shading); zero values indicate no accommodation. Speaker labels (d/n) refer to data collected by different experimenters and do not reflect any differences in speaker characteristics.
• Figure 2: Activation field (top) and memory field (bottom) for strut simulations. The $x$ parameter range has been truncated to highlight small differences in the activation peaks.
• Figure 3: Activation field (top) and memory field (bottom) for bath simulations. The $x$ parameter range has been truncated to highlight small differences in the activation peaks.