The Voice of Experience: Causal Inference in Phonotactic Adaptation

Denby, Thomas

doi:https://doi.org/10.21985/n2-ahda-vs50

Work

The Voice of Experience: Causal Inference in Phonotactic Adaptation

Public

Download PDF

Download All Files (.zip)

Successfully grappling with the widespread linguistic variation of daily life requires speakers to adapt to systematic variation in the environment while discarding incidental variation, based on their prior experience. In the case of phonotactics, speakers’ prior experience is that talkers who differ in their language background are likely to vary in their phonotactic grammars, while talkers who share a language variety are unlikely to do so. As such, we predict that when speakers are exposed to multiple talkers whose phonotactics vary, and those talkers differ in their language background, listeners will infer the variation is systematic and adapt. Conversely, if the talkers share a language background, listeners will infer the variation is incidental, and not adapt. In Study 1, we tested this prediction in a perception experiment, by exposing listeners to two talkers, each of whom exhibited a different phonotactic constraint, in a recognition memory task. In Experiment 1, when listeners were exposed to talkers who differed in their language background (1 English vs. 1 French talker), they showed a high degree of adaptation; when the talkers shared a language background (2 English or 2 French talkers), listeners showed a low-to-moderate degree of adaptation. In Experiment 2, we examined the granularity of listener knowledge of variation in non-native phonotactics by including a novel condition with two non-native talkers (1 Hindi vs. 1 Hungarian talker). Listeners showed a high degree of adaptation even when both talkers were non-native speakers with different language backgrounds, suggesting that listeners make distinctions between different non-native language phonotactics. In Study 2, we examine the role of causal inference in speech production. Recent work suggests adaptation in production may differ from perception, as production may utilize simple associative learning mechanisms that may not take high-level indexical features into account. We explore this question using a modified tongue twister paradigm, in which participants repeat syllable sequences from two model talkers, with each talker exhibiting a different phonotactic constraint. Mirroring Study 1, model talkers either shared a non-native language background; shared a native language background; or differed in their language background (the languages backgrounds in question were German and English). In addition, a control condition was included following previous tongue twister experiments, in which the phonotactic constraint was conditioned on the identity of the adjacent vowel. Results were largely inconclusive—there was some evidence of increased adaptation when participants were exposed to model talkers with different language backgrounds, but the effect was inconsistent. In addition, no effect was found in the control condition. Together, these results suggest that phonotactic adaptation is flexible, but constrained by the causal inferences listeners draw from their prior experience, particularly in perception.

Creator