Modeling the emergence of syllabic structure

Most computational models that have addressed the development of consonant-vowel syllable systems have assumed a preexisting tendency to produce syllabically structured utterances. For instance, the COSMO model assumes a consistent base of jaw movement upon which finer-grained articulatory patterns are learned, motivated by MacNeilage and Davis's frame-then-content theory (FCT). While research operating under this assumption has provided much useful information on infant speech development, it does not address the gradual transition from non-syllabic to syllabic sounds that occurs during the first seven months of human infancy. It is important to consider the role that learning plays in this very significant transition. This paper discusses two computational models that address how infants may learn to produce syllabic utterances. Future work should develop agent-based models of sound production that show how syllabicity itself can emerge in a population of agents.