Models of speech production differ in their claims about syllabificati
on, In a memory-based approach such as that advanced by Dell (1986, 19
88), the syllable structure of each word is stored in the mental lexic
on. In contrast, according to a rule-based approach, planning of speec
h involves the assignment of syllable positions to segments after they
have been retrieved for a word from memory (e.g. Levelt, 1992). Here,
a case is made for the rule-based approach of the WEAVER model of spe
ech production (Roelofs, 1994, in press a). First, I argue that cross-
morpheme and cross-word syllabification point to the need to deal with
flexibility of syllable membership and therefore pose difficulty to a
memory-based approach but not to WEAVER. Secondly, I review empirical
support for the specific form of syllabification realised in WEAVER.
Thirdly, I report a new experiment on syllabification, which supports
WEAVER rather than Dell's model. Finally, the issue of resyllabificati
on is discussed.