Noise and inter‐speaker variability improve distinguishability of auditory, motor and perceptuo‐motor theories of speech perception : An exploratory Bayesian modeling study

International audienceA central issue in speech science concerns the nature of representations and processes involved in communication. Three major sets of theories have been widely argued for and against in this long-standing debate : motor, auditory and perceptuo-motor theories. Arguments have so far mostly been based on experimental data about variability and invariance (coarticulation phenomena supposedly in favor of motor theories ; or motor equivalence principles supposedly in favor of auditory theories ; or co-structuration of the perceptual and motor repertories in the Perception-for-Action-Control Theory, PACT. However, none of these observed isolated properties and their associated arguments are decisive, and the theoretical debate appears to be stagnating. It is our belief that mathematical modeling of these theories could provide breakthroughs in this regard. More precisely, we propose that casting these theories into a single, unified mathematical framework would be the most efficient way of comparing the theories and their properties in a systematic manner. Bayesian modeling is a mathematical framework that precisely allows such comparisons. The trick is that the same tool, namely probabilities, can be used both for defining the models and for comparing them. Such comparisons are more and more widespread in cognitive science ; see for instance the recent works on causal inference and probability matching strategies in multimodal perception, or on theoretical comparison of memory models. Moreover, the use of a unified framework implies that common hypotheses would have common mathematical translations. This also helps toward more principled studies of the competing theories. In our integrative approach, we thus cast the motor, auditory and perceptuo-motor theories into one unifying Bayesian framework in which they all appear as instances of various questions asked to one probabilistic communication model. This allows to compare these theories through quantitative testing in various paradigms