Not only is a generative account more clumsy than one which is solely constraint-based, but using a generative approach obfuscates the real nature of the phenomenon. Since it is the constraints that do the real work in all analyses of syllabification, it is theoretically sound to appeal to a solely constraint-based theory. For example, we account for the distributional fact that word-final consonants are sonorant with the general constraint ALIGN-OBSTRUENT. A generative approach, likewise, would rule out a word such as *[gat], by positing some constraint against obstruents in coda position. However, given that the mechanism of extrasyllabicity is available, the generative analyst must explain why a word-final [t] cannot be considered extrasyllabic.
We account for the distribution of data solely by a small set of ordered constraints: alignment constraints (McCarthy & Prince 1993, Prince & Smolensky 1993), and SONORITY-HIERARCHY, a constraint which formalizes the sonority hierarchy developed in Davis (1990), taking into account both distance and slope of the sonority cline between adjacent segments in order to determine syllabification.
We present the results of a word-game experiment in which native speakers pronounced words and inserted nonsense segments between syllables. Vowel length in stressed syllables indicates whether they were producing open or closed syllables, and their conscious syllabification indicates where syllable boundaries occur. The results fully confirm our analysis. Although there were some productions which did not match our predictions, they can be systematically accounted for by interference from morpheme boundaries or consciously learned hyphenation rules.
Finally, our experiment brought to light a new constraint, ALIGN-NASAL, requiring that nasal segments occur only at syllable edges in Italian, a generalization that has not previously been made in the literature. This accounts for the following observations: (1) initial consonant clusters of the form [stop] + [nasal] are rare, (2) initial consonant clusters of the form [s/z] + [stop] + [nasal] do not occur, (3) medial clusters of the form [sonorant] or [s/z] + [stop] + [nasal] do not occur, and (4) medial clusters of the form [stop] + [nasal] do occur, though infrequently.
McCarthy, John & Alan Prince. (1993) Generalized Alignment. In Geert Booij & Jaap van Marle (eds.), Yearbook of Morphology 1993. Dordrecht: Kluwer. 79-153.
Prince, Alan & Paul Smolensky (1993). Optimality Theory: Constraint Interaction in Generative Theory. Rutgers and U. of Colorado-Boulder ms.