The phonetics and phonology of onset clusters: The case of Modern Hebrew

Abstract

The focus of this work is the phonetics and phonology of word initial bi-consonantal onset clusters. I begin by reporting on a cross-linguistic study of the typological distribution of two features, [sonorant] and [voice], in word initial clusters. There are 4 logical possibilities for combining [?sonorant] (O) and [+sonorant] (S) in bi-consonantal onset clusters, OO, SS, OS and SO, and 16 logical combinations of these clusters. A survey of 62 languages was conducted and it was found that of the 16 logically possible language types, only 4 emerge as occurring: {OS}, {OS, OO}, {OS, OO, SS} and {OS, OO, SS, SO}. The following implications were found: SO -> SS -> OO -> OS. The sub-typologies of OO and SS clusters were examined separately. Next I address the feature [voice] and the distribution of this feature in bi-consonantal onset clusters, focusing on obstruents. Out of the 4 logically possible combinations of [-voice] and [+voice] in obstruent clusters, and 16 possible language types, only 6 emerge as occurring: {[-v][-v]}, {[-v][-v], [+v][+v]}, {[-v][-v], [-v][+v]}, {[-v][-v], [-v][+v], [+v][-v]}, {[-v][-v], [-v][+v], [+v][+v]} and {[-v][-v], [+v][+v], [-v][+v], [+v][-v]} with the following implicational relations: [+v][-v] -> [-v][+v] -> [-v][-v] and [+v][+v] -> [-v][-v] The different typological patterning of the two features implies that it is impossible to predict the typological patterning of clusters of one of these features, based on the other. A language can be of one type in terms of [sonorant] but of a different type in terms of [voice]. The typological patterning of clusters based on the feature [sonorant] does not provide clues about the phonological patterning of the feature [voice]. In particular, I show that languages treat [+v][-v] and SO clusters differently and argue that solutions that have been proposed for the special phonological representation of SO clusters such as appendix, cannot account for [+v][-v] clusters. In the second part, I present an acoustic phonetic study of word initial clusters with different laryngeal specifications in Modern Hebrew. I show that all four types of voicing clusters are realized phonetically. I further show that voicelessness is not always an underlying target, but can be a result of voicing failure in an unfavorable phonetic context.