The purpose of this presentation is to discuss the claim made by Kaye (1995), viz. that non-analytic (also called synthetic) concatenations exhibit the same phonological behaviour as monomorphemic ones. The validity of the Kayean division has already been called into question with regard to languages such as Hungarian, which have a rich and complex morphology. It will be shown that this claim is difficult to maintain even in the case of a language like English (with a relatively impoverished morphology), drawing on examples from Modern and Middle English. At the same time, a working solution will be offered.
In all versions of Government Phonology since 1990, there has been an — explicit or implicit — distinction between two types of lateral forces: those that decide which clusters are well-formed (Constituent Government, Interconstituent Government, Government-Licensing in the 1990 model), and those that produce segmental effects such as vowel-zero alternations, lenition etc. (Proper Government in the 1990 model). The Coda Mirror (Ségéral and Scheer 2001, 2005, Scheer 2004:§110) has set out to unify both, that is to build a system where cluster well-formedness and segmental effects are controlled by just two lateral forces, government and licensing. This ambition may have struck beyond the mark. In any event, even in this system a version of Government-Licensing was left whose only purpose is to control branching onsets (the R of a TR cluster must be licensed by its nucleus); Scheer (2004:§149) has identified its alien character in the system: Government-Licensing does not produce any segmental effect.
Cyran's (2003) theory only concerns cluster well-formedness (even though Cyran tries to maintain the unity of cluster-licensing with alternation-relevant lateral forces). If the ambition to boil down the set of lateral relations to just government and licensing is abandoned, the old division of labour could solve the problem that we are left with at the end of this article: Cyran's licensing is only about cluster well-formedness — call it cluster-licensing; the lateral forces of the Coda Mirror are only about the segmental expression of syllabic constituents: Ségéral and Scheer's licensing may be called segment-licensing (while Government does not need to be renamed). Lateral actorship of nuclear categories, then, is defined at both levels: whether a constituent can govern and seg-license or not is calculated as before in the Coda Mirror; whether it can clu-license is decided independently of its being governed or not.
Selkup is one of the Samoyedic languages that make up one branch of the Uralic language family (Finno-Ugric languages being the other major branch). Although it has been under investigation for well over a century, a good phonological account is still missing. (The same goes for other linguistic fields, of course!)
This talk presents work in progress and therefore does not give a phonological account of Selkup, but is here to show off some of the - in our opinion - more interesting problems of the language's phonology and thus whet other people's appetites. (Of course we also hope Your input will help us to solve some of the mysteries!)