Optimality "Theory" as a Relational Network TheorEM

   Good work has been done in Optimality Theory (OT), one of the newer
branches of generative phonology. Milnes and Wiltshire 2000 (MW)
provide an OT description of the phonology of Russian accentology. The
present study demonstrates that a generalized relational network (RN)
description of the same data not only predicts the correct accent
placement but also the OT constraints and their ordering. Accent
placement, the constraints, and their relative order emerge from the
network. OT is thus derivable from an RN theory.

MW is based on a Laszlo-Ivsic (LI) description of the accent patterns
in the Russian a-stem declension as seen in kajma 'edge' (word-final
accent throughout), zaba 'toad' (stem accent throughout), and zena
'wife', zima 'winter', and ruka 'hand, arm' (all with movable accent).
An LI description assumes that morphemes are marked for simultaneous
stress, for stress on the following morpheme, for stress on the
preceding morpheme (in derivation only), or are unmarked for stress.
Morphemic words, which are the input to the phonology, may therefore
be marked for zero, one, or more stresses. The phonology then sorts
out the stresses, picks the correct one, and deletes the others. MW
use a series of OT constraints, some established, some modified. The
RN description is based on the traditional formulation that Russian
stress is free and phonemic. (It is also called "mobile," but that
refers to morphological paradigms and is handled by a variant of the
LI description.) The RN description says 1) that a phonological word
in Russian has 0-m unaccented syllables followed by one accented
syllable followed by 0-n unaccented syllables, and 2) that any given
occurrence of morphological accent is realized as either a
phonological accent or as zero. This predicts the correct accents, and
a minor modification of the network provides a phonological accent for
morphological words with no stress. Certain details of the network
hierarchy are shown to correspond to the OT constraints and to predict
the hierarchy between them given in MW. This can be done for OT
descriptions of different phonological phenomena in unrelated
languages (e.g. Korean coda simplification). Thus OT descriptions do
not require the assumptions of generative phonology and OT itself is
derivable from an RN theory. It is, in short, a theorEM of a RN
theorY.

Milnes, Irina, and Caroline Wiltshire. 2000. Optimal stress patterns
   in Russian nouns. Annual workshop on formal approaches to Slavic
   linguistics: the Philadelphia meeting 1999, Tracy King and Irina
   Sekerina (eds), 296-313. Ann Arbor: Michigan Slavic Publications