Self-organization of Syllable Structure: a Coupled Oscillator Model

It has been generally claimed that every language has syllables with onsets (CV structure), while languages may or may not allow coda consonants (VC structure). While recent work on Arrernte (Breen & Pensalfini, 1999) has cast doubt on the absolute universality of onsets, it is clear that there is a significant cross-linguistic preference for CV structure (Clements & Key-ser, 1983; Clements, 1990). In addition, evidence from phonological development shows that CV structure is typically acquired before VC for CV structure in distribution and acquisition has been understood as due to universal grammar (UG: Chomsky, 1965) where CV is the unmarked core syllable structure. Yet, the UG hypothesis does not answer to the question " Why is CV the most unmarked structure? " This study aims to provide a rationale, grounded in dynamical systems theory, for why CV is favored across languages by simulating the self-organization of syllable structure in phonological development using a model in which syllable structures are defined by the coupling graph in a system of gestural planning oscillators that control patterns of intergestural relative timing. The simulation shows that, due to the hypothesized stronger coupling inherent in CV compared to graphs, CVs emerge earlier than VC graph. The model of syllable structure based on coupled planning oscillators (see section 2 below) has been developed to account for a variety of empirical observations about speech production (Nam & Saltzman, 2003; Goldstein et al., 2006; Nam et al, submitted a, b), independently of any consideration of acquisition facts. In addition to the explanatory weakness of the UG hypothesis, two additional empirical observations about phonological development are not easily accommodated by the UG hypothesis. One is that the delay in the emergence of coda consonants varies across languages as a function of the