Two proofs that classic Optimality Theory is expressively weaker than ordered rewrite rules
Brian Buccola
December 2012

This paper provides the first mathematical demonstration that classic optimality-theoretic grammars, consisting of just two levels of representation (underlying and surface) and of just two types of constraints (markedness and basic faithfulness), are formally expressively weaker than ordered rewrite rules. Specifically, it is shown that there are sets of patterns which can be expressed by ordered rewrite rules but which in a two-level OT grammar require at least one faithfulness constraint that assigns a violation mark to certain pairs of input-output segments in correspondence, but only when they occur in certain contexts. Such a faithfulness constraint, however, which references segments other than the correspondents under consideration, is beyond the limits of what is standardly and intuitively considered classic OT. In addition, at least one such set of patterns (Canadian raising) is attested in natural language, thus making classic OT, as defined, and any more restrictive variant thereof, empirically inadequately expressive. The paper ends with a discussion of how classic OT might be extended slightly so that it becomes empirically adequately expressive while still being provably weaker than ordered rewrite rules.
Format: [ pdf ]
Reference: lingbuzz/003038
(please use that when you cite this article)
Published in: Second evaluation paper, McGill University (NB: This paper is mostly superseded by Buccola 2013, "On the expressivity of Optimality Theory versus ordered rewrite rules", Formal Grammar 2012/2013)
keywords: optimality theory, rewrite rules, computational phonology, finite-state phonology, phonological opacity, phonology
Downloaded:101 times


[ edit this article | back to article list ]