| Summary: | Autosegmentalism invariably represents geminates in a symmetrical one-to-many relationship — as feature bundles or root nodes attached to two structural units: x-slots, moras, or C-slots. This symmetry, however, is often not reflected in their diachronic origin. For instance, in Blevins’ (2008) Type 1 pathway, only the second C of a consonant cluster (CC) ever determines the geminate: CxCy > CyCy, *CxCx (e.g. Latin > Italian). Moreover, although most synchronic processes identify geminates as symmetrical, there is an exception: geminate integrity. Unlike CCs and long vowels (LVs), geminates never ‘break’ by epenthesis: *CyCy > CyVCy. We propose that this is central to understanding the true nature of geminates, and present analyses in two frameworks. The first is ‘control by contiguity’, which uses head-dependent ‘control chains’ (Russo 2013). A control relation applies between a specified and an unspecified position: -C. Inalterability and integrity result from the asymmetry of the geminate’s positions. The second is based on Strict CV. This restricts a geminate’s melody to one of its two skeletal positions. Unlike CC and LVs, geminates do not involve a ‘trapped’ empty V position that could host epenthesis and cause breaking; the difference between LVs and geminates follows from framework-internal forces and suggests that melodic branching always requires licensing. These two approaches share the insight that the representation of geminates is not symmetrical, like that of long vowels.
|
|---|
