Abstract :
[en] The general framework of this research is the emergence and the collective structuring of phonological systems. In this study, we try to identify some of the mechanisms involved in the propagation of sound changes within a speech community. Our main hypothesis is that sound systems result from mimetic interactions between individuals. Mimesis is defined as a supramodal, motor-modelling skill which creates representations that are retrievable from memory. Its function is to represent events in a conscious, intentional and deliberate way (Donald, 1991). In the context of phonology, mimesis is considered as the competence to develop and amplify variation.
This study is an attempt to implement sound change in the laboratory. The aim is to give rise to modifications of phonetic realizations from speakers exposed to a simple interactive situation. The experimental setting is the following: ideograms are presented on a computer screen. The computer randomly selects one of S1, S2, S3 to name each ideogram, where S1, S2, S3 can be either a speaker or a recorded speaker.
Four subjects have been selected from two dialectal groups of the French-speaking Belgian community (two per dialect, A1, A2, and B1, B2). Prior to the experiment, every individual learned to associate words with ideograms. In a pre-test, all subjects are recorded individually as they pronounce learned words when prompted with corresponding ideograms. The recordings are used as subjects Reference Pronunciations (RP).
Two experiments will be run, each consisting in four one-hour sessions, as described in Table 1. In the first experiment, the ideograms are named alternatively by the individual and by prerecorded RP. The productions of the individual are recorded for analysis. This design allows to study the potential evolution of an individual's realizations, when exposed to a dialect different from his own. In the second experiment, the ideograms are named alternatively by three individuals. Here people are involved in a more interactive situation.
The corpus consists in twenty items selected to study two well-documented dialectal variants within the French-speaking Belgian community.
The distance between two productions is obtained by dynamic programming to allow global comparisons. The acoustic cues are the classical MFCC computed every 5ms. This approach allows us to quantify the evolution of the realizations of a speaker in a given experiment with regards to his own and to the others' RP.
These experiments are intended to favour the mimetic adaptation of speakers, with the aim of getting insight on the mechanisms involved in the propagation of phonetic changes.
