Avoidance of final monopositional vowels: Evidence from the k-dialects of Yemen

A group of Arabic dialects in Yemen exhibit a velar k in the subject suffixes of the perfect paradigm. The 1sg subject suffix surfaces in the various dialects as -ku, -k or -kw. In addition, the vocalization of the base may or may not be colored with a round vocalic quality, depending on both the realization of the suffix and the verbal type fa'al or fi'il. Based on inquiries among speakers from Lower Yemen, we propose a path of evolution that leads from -ku to the labialized -kw, to a "colored"stem, and finally to the grammaticalization of coloration and loss of labialization. Two pressures propel the passage between stages: a functional pressure to distinguish between 1sg and 2msg, and a phonological pressure to avoid monopositional final vowels. The phonological pressure is shown to also motivate palatalization in the 2fsg -ki → -ky (→ -š), as well as the effect of pre-pausal nasal insertion, viz. -ku → -kum# and -ki → -kin#. We further show that final vowels resulting from the interaction of the subject suffixes with object clitics are phonologically long - even if phonetically neutralized - and hence they do not violate the phonological requirement. The formal theories of strict CV (Lowenstamm 1996, Scheer 2004) and Element Theory (Kaye et al. 1985) are used to explain the asynchronized development in the different verbal patterns, as well as the extent of the phonological ban on monopositional vowels.