Underlying Representation and Acoustic Analyses of Two-Consonant Initial Clusters and Prothesized Vowels in Najdi Arabic

Abstract

Classical Arabic (CA) does not allow clusters in the onset (Kiparsky, 2003), so the traditional wisdom has been that Arabic in general disallows onset clusters (Akashah, 2007). Nonetheless, scholars have found that some dialects of Arabic do allow clusters in the onset, such as the Arabic spoken in Damascus (Cowell, 1964), Morocco (Boudlal, 2001), Amman (Daana, 2009), and the Najd region of Saudi Arabia (Abboud, 1979). In three experiments, the present study investigates the underlying representation of two-consonant word-initial clusters and how they surface in Najdi Arabic (NA). It contributes to the continuing debate in the literature regarding the relative sonority of obstruents, the limited existing literature on Arabic vowels (lexical vs. non-lexical) in terms of perception and production, and proposals claiming that syllabification can be determined by measuring the temporal organization of segments. The first experiment examined production, seeking to determine which consonant sequences were acceptable to speakers of NA using a constrained set of onset clusters with three sonority contours: falling, plateau, and rising. Participants tended to produce rising sonority clusters as [C1C2] and rarely as [VC1C2] but produce plateau and falling sonority clusters mostly as [VC1C2] and rarely as [C1C2], [C1VC2], or [C2]. Comparing the vowels produced in [VC1C2] rising, plateau, and falling sequences revealed durational differences; the vowels added to rising sonority clusters appeared to be shorter and less intense than those added to falling and plateau sonority clusters. These results suggest sonority plays a role in NA word-initial cluster production. The second experiment had two phases, encompassing production and perception. The first phase tested the production of lexical vowels by NA speakers in the same environment as the vowels found to surface in Experiment 1 (__C1C2). The aim was to check whether the vowel in Experiment 1 was similar in duration and intensity to the lexical vowel. Results indicated that lexical vowels were significantly longer in duration and higher in intensity than the vowels in Experiment 1 in all three sonority contours. The second phase tested participants’ perception of these different vowels and whether it matched their production. The task was to choose the optimal spelling between #VC1C2X, #C1VC2X, and #C1C2X while hearing nonce words starting with a vowel of lexical or non-lexical length, in addition to a third category where the whole vowel was deleted. The results exhibited a difference of perception between these categories; the spelling #VC1C2X was chosen more when hearing a word starting with a vowel of lexical than non-lexical length and more with non-lexical length than with no vowel at all. When hearing nonce words starting without a vowel [#C1C2], participants tended to hear an illusory vowel approximately half of the time. This finding suggests the vowel in front of these consonant sequences is in the NA underlying representation: /VC1C2/. Experiment 3 examined production, seeking to determine the syllabification of NA clusters that emerged in Experiment 1. Participants were asked to read stimuli that consisted of seven pairs (#C1C2VX vs. #C2VX) where the only difference between each pair was the initial consonant added (C1). The goal was to acoustically measure the temporal organization of these pairs to report on consonant syllabification. If they were more stable when measuring the right edge (to some known anchor / nucleus), NA word-initial clusters would best be syllabified as simplex (#C1.C2VX). If, however, the center (to the same anchor) for these pairs tended to be more stable than the right edge, these clusters may be better syllabified as complex (#C1C2VX). All pairs tended to be more stable when measuring the right edge to anchor, indicating the underlying represen