The problem of non-optional directional iterativity
Bert Vaux
May 2015 

Dybo 1977 proposed that the core facts of Abkhaz stress can be captured as in (1), on the assumption that vowels can bear lexical accents in underlying representations: (1) Primary surface word stress falls on the leftmost lexically accented vowel that is followed by an unaccented vowel or word edge. If there is no lexical accent in the word, surface stress falls on the last vowel of the word. Spruit, Kathman, and Trigo have shown that Dybo's insights can be derived straightforwardly in Rule-Based Phonology (RBP). In the Halle & Idsardi 1995 formalism, for example, we can derive the basic Abkhaz system via the operations in (2). (2) i. Project a right bracket ) for all lexical accents. ii. Line 0 Edge Marking: LLL iii. Clash Deletion: ) ➝ Ø / _ * ) [iterative, L➝R] iv. Project rightmost element of Line 0 feet to Line 1 v. Project leftmost element of Line 1 feet to Line 2 The conflicting directionality identified by Dybo results from Left vs Right headedness on Lines 0 and 1 respectively (iv, v), and the iterativity and directionality via iii (cf. Howard 1972). On the other hand, no version of Optimality Theory currently on the market is able to account for a system of the Abkhaz type, due to the ability of the Clash Deletion rule to apply iteratively and directionally within individual morphemes. In this talk I present the Abkhaz facts and their analysis in RBP and the leading variants of OT. I show that the Classic OT tenets of globalism/parallelism and minimal violation favor outputs which do the global minimum necessary to avoid stress clash, which harmonically bound the desired winners with their greater number of clash deletions. Versions of OT that allow staged computation sensitive to morphological structure (e.g. Stratal OT or Orgun's Cyclic OT) can deal with at least a subset of the cases where no more than one deletion happens per morpheme, but Strata OT fails with forms involving more than one deletion per stratum and Cyclic OT fails with forms involving more than one deletion per morpheme. Harmonic Serialism (and perhaps OT-CC) have the power to generate the desired outputs, but cannot rule out equally harmonic outputs produced by derivations that do not apply clash deletion in L→R order. The problem in brief is that the directionality of clash deletion has to be built into the process itself, which isn’t allowed in existing forms of OT, as they do not recognize the existence of processes.

Format:	[ pdf ]
Reference:	lingbuzz/007090 (please use that when you cite this article)
Published in:	Manchester Phonology Meeting
keywords:	phonology, stress, optimality theory, abkhaz, iterativity, optionality, meeussen's rule, rule-based phonology, phonology
Downloaded:	481 times

 

[ edit this article | back to article list ]