Sonority and Other Constraints in Gitksan Consonant Clusters
Jason Brown [+]
University of Auckland
While many languages of the world are very limited in what kind of consonant clusters are tolerated by their phonologies, many other languages allow extensive clustering of consonants. Gitksan, an Interior Tsimshianic language of British Columbia, Canada, allows relatively extensive clustering of consonants. This paper outlines the phonotactics of Gitksan (Tsimshianic) by attempting a summary of the generalizations presented in Rigsby (1986). Word-initial position, which is the focus of the present study, yields many combinations of possible consonant sequences in the language. For example, in word-initial bi-consonantal clusters, stops can be sequenced before fricatives, and they can even be sequenced before other stops. Fricatives exhibit a similar distribution, where they can be sequenced before both stops and other fricatives. Thus, there is prima facie evidence that any constraints on sonority sequencing in the language are lowly ranked. Curiously, however, there are relatively severe restrictions on sonorants. Sonorants cannot co-occur, and while sonorant consonants can be sequenced after fricatives (i.e. fricative + sonorant) in word-initial position, there is a gap corresponding to stop + sonorant sequences. This gap extends beyond word-initial position: there are no stop + sonorant sequences present in the language (i.e. this type of sequence is missing in word-initial, word-medial, and word-final positions, regardless of syllable affiliation of the consonants). This gap is unexpected, since fricative + sonorant sequences are tolerated by the language. A sonority account (where sonority must rise a given amount in an onset) is inadequate, as fricatives are presumably more sonorous than stops; likewise, different versions of sonority, such as the Syllable Contact Law are likewise inadequate, as the restriction is evident in word-initial contexts where the consonants are tautosyllabic. An alternative approach to sonority is entertained for this gap, namely, one based in perceptual similarity (cf. Henke, Kaisse & Wright 2012), where it is the perceptual distance, rather than sonority distance, that is encoded in constraints on clustering.