On the ambiguous segmental status of nasals in homorganic NC sequences

A fundamental problem in the analysis of pre-nasalized segments is to determine whether these homorganic nasal-obstruent (NC) sequences are, in fact, best considered unit segments or clusters. The segmental status of NC sequences is hard to resolve because, as work like Maddieson & Ladefoged (ML Herbert 1975; Hyman 1992; Katamba 1985; Tucker 1962) (a) -mala ‘finish’vs. -maala ‘smear’ (b) –taanda ‘betray’ (*-tanda) (c) mu-temuzi‘murderer (cl.1)’ vs muu-mbowa ‘executioner (cl.1)’ Since work like Clements (1986), pre-NCL in Bantu languages has been analyzed as a form of compensatory lengthening. The NC sequence is a cluster in the input (2a). As Bantu languages are said to allow only open syllables, the NC sequence must be syllabified as a (non-moraic) Onset. To avoid violating the Sonority Sequencing Principle, the NC must become a unitary pre-nasalized segment on the surface (2b). The vowel lengthens to compensate for the loss of the segmental status of the nasal, taking over its timing unit (Clements 1986; Hubbard 1995; Hyman 1992; Schadeberg 1991; van de Weijer 1996; etc.): (2) (a) Input to lengthening (b) Output μ μ μ μ σ