A Stratal OT Account of Word Stress in the Mehri of Bit Thuwar

In this paper, we provide a synchronic account of word stress in the Modern South Arabian language, Mehri, as spoken by members of the Bit Thuwar tribe. The data is taken from the first author’s own fieldwork working in Central Dhofar with members of the Bit Thuwar sub-tribes–Bit Iqhōr in Rabkut and parts of the mountains that receive the monsoon rains, and Bit Āmawsh in Dhahbun–with reference, where appropriate, to Johnstone (1987). This paper is a significant expansion and a partial revision of the short discussion on word stress in Watson (2012: 34–35).

We begin with a brief background to Mehri within Modern South Arabian. We then discuss word stress patterns in Mehri, following Hayes’ (1995) metrical stress theory; here we show that Mehri is a head-first or trochaic language, namely that in (Cv)CvCv(C) forms, stress is placed on the left-most syllable. We show that in contrast to Arabic trochaic dialects, the domain of stress is the stem and the stem with stem-level suffixes rather than the entire prosodic word. The implications for this is that Mehri exhibits an opacity with regard to stress as seen in Arabic dialects in which three consonant clusters receive vowel epenthesis after the left-most consonant, as in: šuft-ha > šufitha ‘she said it f.’. (These are termed vC-dialects in Kiparsky 2003, so called due to the position of the epenthetic vowel in relation to the medial consonant in a sequence of three). By opacity, we mean that word stress is not assigned as would be predicted by the stress algorithm.

In contrast to Arabic vC-dialects, however, opacity is due not to the interaction of epenthesis and syncope (cf. Kiparsky 2003), but rather to the lack of visibility of word-level suffixes to stress. In line with van Oostendorp’s (2002) analysis of unstressable suffixes in Dutch, we argue that word-level suffixes are invisible to stress because they are not incorporated into the prosodic word, but rather adjoined to it. The adoption of Kiparsky’s Stratal Optimality Theory approach enables us to capture Mehri stress assignment succinctly: stress is assigned at the stem-level according to weight and position, and suffixation of word-level suffixes can no longer affect stress assignment due to the high ranking of STRESSIDENT (Collie 2007), which requires stress to remain on the stressed syllable of the stem, and the low ranking of *ADJOIN, which mitigates against the adjunction of affixes to the prosodic word. We also show that Mehri exhibits limited lexical stress, and suggest that attempts to account for stress in these cases in terms of a quantity model on the part of many researchers, including the main author of the current paper, has led to the incorrect transcription and interpretation of these elements.