Vowels phonemic depending on the part-of-speech

[Ser]

When we talk about epenthetic vowels, we usually mean to say they're inserted because of the particular structure of a word. But if the location of a vowel can be predicted taking on account the structure AND the part-of-speech of a word, can we still call it epenthetic? Or is the part-of-speech too much for epenthesis?

Here's some examples from Moroccan Arabic.

• [ħbəs] 'to imprison (sb), imprison!'
  [ħəps] 'jail'
  (Note that [p] is the normal allophone of /b/ before a voiceless consonant.)
  
  [frˤəħ] 'to be happy, be happy!'
  [fərˤħ] 'happiness'

One way to analyze these pairs is that /CCC/ is [CCəC] if it's a verb but [CəCC] if it's a noun. So basically we get /CCC/^noun and /CCC/^verb. This is what Ernest Abdel-Massih does at least. Would you agree? Or is better referred to as a phonemic vowel, i.e. /ə/ and therefore /CCəC/ and /CəCC/?

[clawgrip]

I don't think the vowel here is epenthetic; it appears to be a morpheme that modifies the meaning of the root, rather than simply a sound that is inserted to make pronunciation easier. In this case it is acting as an infix (or transfix). It seems to be contrasting Ø-ə (verbal) with ə-Ø (nominal).

[spats]

Agree, and it doesn't matter where the vowel came from. It has become phonemic.

This happens all the time. Some feature is purely allophonic, but ends up following a pattern because some other language feature changes from one word form to another. For example, perhaps one form has a prefix and the other a suffix, or the stress moves, or another vowel is long in one form and short in another.

Then, due to sound change or grammar change, the sound causing the influence disappears. Following the disappearance, the previously allophonic distinction becomes phonemic.

Most if not all sound splits happen this way. In this case, the sound that split was simply Ø!

[Astraios]

[ħbəs] 'to imprison (sb), imprison!
[ħəps] 'jail'

[frˤəħ] 'to be happy, be happy!'
[fərˤħ] 'happiness'

Also compare Maltese aħbes (doesn't seem to be in use anymore, but) vs. ħabs and ifraħ vs. ferħ, and since Maltese hasn't lost vowels to the same extent as Moroccan Arabic, I'd go with it being just a phonemic vowel.

[Ser]

I don't think the vowel here is epenthetic; it appears to be a morpheme that modifies the meaning of the root, rather than simply a sound that is inserted to make pronunciation easier. In this case it is acting as an infix (or transfix). It seems to be contrasting Ø-ə (verbal) with ə-Ø (nominal).

Yeah, this would seem to me to be the better analysis.

As for Abdel-Massih, it's also annoying that in his course for Moroccan Arabic he only rarely marks the "epenthetic" schwas and stress. It may be obvious to him but not to a non-native! You see "sl:m" and it isn't clear if it's ['səl.ləm] or [səl.'ləm].

Most if not all sound splits happen this way. In this case, the sound that split was simply Ø!

I'm not sure about that, Moroccan schwas are often vowels in other dialects. Just look at the examples provided by Astraios, where Moroccan [ħbəs] patterns with Maltese aħbes (and Standard Arabic 'ħabasa/'ʔiħbis/ħbis); and [ħəps] patterns with ħabs (and Standard Arabic ħabs). Although as you can see with the former group, if there's only one schwa it doesn't always correspond to the stressed vowel of the cognate word in Standard Arabic or Maltese (or an Arabic dialect from the east of Morocco).

[kodé]

When we talk about epenthetic vowels, we usually mean to say they're inserted because of the particular structure of a word. But if the location of a vowel can be predicted taking on account the structure AND the part-of-speech of a word, can we still call it epenthetic? Or is the part-of-speech too much for epenthesis?

Here's some examples from Moroccan Arabic.

• [ħbəs] 'to imprison (sb), imprison!'
  [ħəps] 'jail'
  (Note that [p] is the normal allophone of /b/ before a voiceless consonant.)
  
  [frˤəħ] 'to be happy, be happy!'
  [fərˤħ] 'happiness'

One way to analyze these pairs is that /CCC/ is [CCəC] if it's a verb but [CəCC] if it's a noun. So basically we get /CCC/^noun and /CCC/^verb. This is what Ernest Abdel-Massih does at least. Would you agree? Or is better referred to as a phonemic vowel, i.e. /ə/ and therefore /CCəC/ and /CəCC/?

Coincidentally enough, I just went over an analysis of this data in my Prosodic Phonology class. Basically, verbs get a different epenthesis pattern than nouns in order to agree with other forms of the verbal paradigm, which have to epenthesize in that way because of their subject suffixes. This paradigmatic agreement outweighs the normal application of the phonology, which prefers the epenthesis patterns in nouns.

Noun phonology and verb phonology are often different. English stress differs between nouns and verbs (hence the famous noun-verb pairs like ['ob.ject] (noun) and [ob.'ject] (verb). The (comparatively) well-known long-vowel lowering rule in Yokuts languages is only active in (strong) verb roots; long vowels do not get lowered in nouns (or loanwords).