Brief note on a synchronic paradox in Levantine Arabic phonology: word-final long vowels are pronounced short unless there is a suffix in which case they're realized as long: /ramaː/ > [ˈ http://ra.ma ] 'he threw' but /ramaː/+/haː/ ‘3fs’ > [ra.ˈmaː.ha] 'he threw it.’

Vowel length plays an important role in stress assignment. So a long vowel in an open final syllable realized as a short vowel wouldn’t receive stress. However, a long vowel in a closed final syllable would: /ʃabiːh/ > [ʃa.ˈbiːh] ‘similar.’

At some point in LA, the 3ms object suffix had a form like *-uh, with a variant *-h after words that end in long vowels, like /ramaː/. Thus /katab/+/uh/ > [ˈka.ta.buh] ‘he wrote it’ and /ramaː/+/uh/ > [ra.ˈmaːh] ‘he threw it.’

However, as part of a general trend in LA, aspiration in this morpheme has been lost (in northern LA this trend is more advanced and affects the 3fs suffix /haː/ as well). So the new 3ms suffix is /u/ or /o/ with a null variant after long vowels: [ˈka.ta.bu] and [ra.ˈmaː].

But [ra.ˈmaː] violates LA stress rules on long vowels in open final syllables. So synchronically, depending on how you look at it, LA either has a null consonant that solely exists in this one allomorph or there is an h-deletion rule for an h-sound that is entirely abstract.

But perhaps I’m overcomplicating matters. If anyone has a better explanation for [ra.ˈmaː], please share!