The dynamics of multilingualism

Levelt’s blueprint of speech production and perception (Levelt 1989; Levelt, Roelofs, & Meyer 1999) has been widely accepted and is well established. One of the crucial characteristics of the model is its modularity, which is required to account for the speed with which we speak (about two to three words per second). At the same time, the modular principle of the model is also its Achilles heel. In current dynamic approaches to language processing, it is assumed that the language system is not stable and static, but instable and dynamic. In these approaches, language production and comprehension are not seen as sequential processing in a series of separate modules, but as an integrated dynamic system that is characterised by a multidimensional state space with numerous attractors at different levels. In the wake of earlier suggestions to adapt the Levelt model to the multilingual speaker (De Bot 1992; De Bot 2000), this paper discusses the robustness of the model to a dynamic language system and tests to what extent it is in line with a dynamic, usage based approach to language processing.