Adyghe syllable structure: From empirical data to generalizations

This paper describes the syllable structure of Adyghe, a Northwestern Caucasian language spoken in Russia, Turkey and some other countries. The range of possible syllable structures in West Circassian varies from simple structures such as V or CV to complex ones such as CCCVCC or CVCCC. I discovered а considerable variability in interspeaker syllabification judgements during а preliminary experiment. In order to create a generalized model of Adyghe syllable structure I used only word-initial syllable onsets and word-final codas taken from words in Yu. A. Tkharkakho’s Adyghe-Russian dictionary. I disregard word-medial consonant clusters because of unclear syllabification rules. I found that the structures of onsets and codas in Adyghe differ, but nearly all attested consonantal classes can occur both in onset and coda positions. Fricatives and plosives show similar distributional patterns, while sonorants differ from obstruents in the following: (a) there are no sonorant clusters; (b) sonorants always occur alone in onsets, so the only type of onset clusters consists of obstruents; (c) there are no syllables with the structure “vowel-sonorant-obstruent”, but there are syllables with the structure “vowel-obstruent-obstruent”; (d) complex codas of three segments can only have the structure “sonorant-obstruent-obstruent”. I also calculated expected cluster proportions on the assumption that the occurrences of all consonant classes in initial or final position are independent events (i. e. there are no mutual contextual restriсtions). Having compared expected and actually observed proportions, I discovered the following: (a) clusters with obstruents are found more often than expected; (b) clusters with sonorants are less frequent than expected; (c) clusters appear more often than expected. These observations allow us to reject the null hypothesis that there are no mutual contextual restrictions on consonant types in initial or final positions.