I agree with Imralu that it would be nice to have a separate thread for this topic. If a mod agrees, maybe some of the old posts can be merged into this thread.
Here is my summary of what I understand at the moment.
- Some people believe in the concept of "ambisyllabicity" (that a consonant can belong to two syllables simultaneously) and favor analyzing the middle consonant of words like "letter", that have a single intervocalic consonant after a stressed short vowel, as ambisyllabic. I haven't found much support for this concept in the linguistic literature I've read, though.
- Some people believe in the principle of maximizing onsets. It seems that valid onsets are generally defined by looking at what consonant clusters can occur at the start of words.
- Some people believe the constraint against word-final stressed short vowels in English is part of a wider constraint that forbids short vowels without any following coda consonants in stressed syllables. This implies the syllabification /ˈlɛ.tər/ is not possible, since /ˈlɛ./ would not be accepted as a valid syllable. As far as I know, everyone agrees that short vowels can occur without following coda consonants in unstressed syllables: the most obvious examples of this are reduced vowels and word-final /ə/ (also word-final /ɪ/ in accents where the "happy" vowel is not tense), and there are also words like "tattoo" where if we reject ambisyllabicity, it seems clear we must divide the syllables as /tæˈtuː/.
- John Wells in particular not only adopts this analysis, but goes even further and has argued that consonants in English are preferentially syllabified with stressed syllables, subject to most of the usual restrictions on valid codas and onsets, so he would not only have /ˈlɛt.ər/ for "letter" but also /ˈbiːt.ər/ for "beater". Some of his arguments can be viewed online on the following web page: "Syllabification and allophony"
Some random things I've been wondering about:
- how to divide "transfer". Morphologically it's trans.fer, but I think I divide transport as tran.sport (the /p/ is definitely not aspirated) so morphology doesn't seem to be relevant. It's clear /sf/ is a possible initial consonant cluster in modern English; pretty much everyone has it in "sphere" etc, and I believe I also assimilate "sv" in foreign names like "Sven" to /sf/. I can't perceive nts-epenthesis in my pronunciation of this word, but I also don't really perceive it in my pronunciation of "trance" or "dance". Then again, for some reason I kind of feel like my prints-prince merger is more a matter of deleting the [t] in "prints" rather than adding a [t] to "prince". I do definitely feel like I have mps-epenthesis though in words like "glimpse", though that might just be an artifact of the spelling.
- Related to that, what about "hamster"? I think I can have epenthetic p here, but also the /t/ is clearly not aspiratable for me which I think in Well's system would count as evidence that it is in the same syllable as the preceding /s/. If plosive epenthesis between nasals and fricatives only applies within syllables (as Wells argues) this means the word should be syllabified as /hæm(p)st.ər/, which doesn't seem right to me at all. I guess I would be inclined to conclude from this that Wells is wrong and this type of epenthesis can occur between syllables.