> David Bell wrote:
> >Don't apologize, until this discussion came up, I hadn't even
> realized the
> >complexity involved here myself.  I have, however been pronouncing this
> >conlang with essentially the same phonology and stress patterns
> for a good
> >number of years now and so I'm very certain of my intuitions
> here.  A lot of
> >other things about the language have changed over the years, but
> these have
> >been fairly stable.  Just what universal tendencies of syllabification do
> >you feel are being violated here?  When I read amman iar, its
> sounds quite
> >natural to me.
> Some linguists have argued that syllabification tends to follow the
> following set of constraints:
> -- Avoid vowel-initial syllables whenever possible
> -- Avoid consonant-final syllables whenever possible
> -- Avoid clusters of two or more consonants within the same syllable
> I've ranked these constraints in order of inviolability (based on cross-
> linguistic tendencies).  These constraints would produce the following
> syllabification patterns, which seem to be valid for a large number of
> natlangs that I'm aware of:
> (1)  VCV is almost always syllabified as V.CV, not VC.V
> (2)  VCCV is typically syllabified as VC.CV, not V.CCV or VCC.V
> (3)  VCCCV is typically syllabified as VC.CCV or VCC.CV, not
> Although (2) and (3) are violated in many languages, some linguists
> have argued that (1) is a strong tendency - a near universal, even.

Fascinating!  I haven't seen this before.  Do you have some references that
I could pursue?

Alas, amman iar does violate (1) and (2) quite regularly.

> Of course, the question of how to syllabify consonants is complicated
> in some languages (like English) which display "ambisyllabicity"
> effects.  In certain VCV sequences in English, native speaker intuitions
> suggest that the consonant actually belongs to both syllables at once.
> For example, when asked to break up the word "happy" [haepi] into
> syllables, many native speakers will hesitate between [hae.pi] and
> [haep.i].  When asked to pronounce these words slowly, syllable
> by syllable, they will often come up with [haep...pi], repeating the
> consonant as both coda and onset.

This I think might occur with geminate consonants, thus the hypothesized
*erinnis could be either er.inn'is or'nis, I think.

> Perhaps (he said, brandishing Occam's razor one final time) Amman
> Iar has ambisyllabicity as well?  Perhaps the "n" in "erinis" is really
> both syllable-final and syllable-initial?  Just a shot in the dark...

As appealing as that razor sharp analysis is, I'm afraid not.  While
*erinnis could be ambisyllabic, erinis most certainly is not.

However, I still prefer that part of your analysis of case endings in which
the stress shift is primary and the gemination concomitant.  I really only
requires a slight change.  Case endings trigger a shift in stress to the
final syllable of the root (erinis[er'inis]+en > *erinisen[erinis'en]).
amman iar metrical rules prohibit stress on a syllable with a simple vowel
followed by a single consonant, and so the final consonant of the root
undergoes gemination (> erinissen[eriniss'en] or ambisyllabically

Still a clean little application of ole Occam's tool.