The Coda Mirror v2 (but not exactly 2.0)

… Studies (Journal of the Phonological Society of …, 2007

There is a well-established asymmetry in the behavior of medial consonant clusters: the first consonant in the cluster can undergo assimilation or deletion, but the second consonant in the cluster cannot. This article presents an explanation for that asymmetry based on a version of Optimality Theory with candidate chains ). The key idea is that a consonant can only assimilate or delete if it first loses its place features by debuccalizing, and debuccalization is only possible in coda position.

2018

1983

The Linguistic Review, 2000

In this article I firstly propose a general framework for formulating interconstituent relations that either 'license' or 'govern' the occurrence of emptyheaded or branching constituents. The Government Phonology literature has put forth a variety of such relations, with different terminology being used by different authors. Here I suggest that the common function of all these interconstituent relations (which I simply all call interconstituent licensing constraints) is to control the distribution of 'marked' syllabic constituents (onsets and rhymes), where by 'marked' I refer to constituents that are empty (−1) or branching (+1), both deviating from the unmarked constituent that contains exactly one (1) segment. Allowing for some parametric variation, I show that each marked constituent must be licensed by the immediate following constituent as well as by the following constituent of the same type on the relevant projection: O←R, R←O, R←R, O←O. In all cases the licensor (to the right of the arrow; all relations are right-headed) cannot be empty-headed. Since the licensees are either empty-headed or branching, we arrive at eight types of licensing constraints. The discussion shows that of these eight, relations in which the licensee is an O, especially a branching O, are the least needed. The discussion of interconstituent licensing constraints is concluded by a brief discussion of long vowels (argued to be sequences of two rhymes, the second of which is empty) and a brief discussion of how cross-linguistic differences in the applicability of interconstituent licensing constraints should be handled. A tentative proposal is that OT-style ranking can be understood as a case of 1. I wish to thank Inkie Chung, Nancy Ritter and Tobias Scheer, and two anonymous reviewers for useful comments on an earlier version of this article. My understanding of these issues is also due to the discussion in a workshop on licensing in Vienna (October 31, 2002) where a rudimentary version of this paper was presented.

2002

In this paper, I argue for a bifurcation of function words in standard Serbian into the free and bound class. I further argue that, while the properties of the free class are to be captured by a general set of prosodic constraints, which account for both lexical elements<br> and free functional elements, the status of bound function words, or clitics, calls for a different approach. If constraints were asked to distinguish between the two classes of function words, this would obviously call for class specific constraints. My proposal is to prespecify clitics as prosodic affixes, associating their lexical entries with prosodic subcategorization frames. The full burden of the distinction between free and bound function words is thus placed on prosody, and implemented through lexical prespecification. This accounts both for cases of distributional overlap between free and bound function words, and those in which distributions of the two classes diverge. In sum, I opt here for pro...