[lingtalks] Linguistics Colloquium today 3/6/06

[Katie McGee]

Just a reminder that there will be a Linguistics Colloquium today at 2:00!

Time: 2 pm
Place:  2148 McGill (TV Studio)
Speaker: Erik Bakovic
Title: Phonological opacity and counterfactual derivation

Abstract:

It has become almost axiomatic that standard parallel OT (SP-OT) is  
ill-suited for the treatment of opaque processes of the form A -->  
B / C_D, both non-surface-true (instances of A in C_D) and non- 
surface-apparent (instances of B derived by P in environments other  
than C_D). In particular, McCarthy (1999) claims that although some  
types of non-surface-true generalizations can be ex-pressed directly  
in SP-OT, non-surface-apparent generalizations cannot be expressed  
without some "refinements" (e.g., McCarthy's Sympathy Theory). In  
this talk I show that there exists an important type of heretofore  
unrecognized non-surface-apparent generalization, which I call  
counterfactual derivation, that can be expressed directly in SP-OT  
without any further refinements.
In Lithuanian, the verbal prefixes /at/ and /ap/ surface with an  
epenthetic vowel -- as [at'i] and [ap'i], respectively, where [C']  
denotes a palatalized consonant -- if the initial consonant of the  
stem is either identical to the prefix-final consonant or differs  
from it only in terms of voicing or palatalization (or both). The  
fact that vowel epenthesis appears to ignore differences in both  
voicing and palatalization is not accidental: the first of two  
adjacent consonants independently assimilates to the second in terms  
of voicing and palatalization in these (and other) contexts. Thus, if  
a vowel were not epenthesized, the expected result after the  
application of assimilation would be a pair of adjacent identical  
consonants in all cases.
Epenthesis should thus ideally be stated to apply only between  
identical consonants, leaving the correct outcome in all cases to be  
determined by the interaction with assimilation. This avoids the  
coincidental duplication of reference to the same features in both  
processes. This generalization in the statement of epenthesis is non- 
surface-apparent, because epenthetic vowels appear in contexts other  
than between completely identical consonants; to wit, they also  
appear between consonants that differ in terms of one or both of the  
features that assimilate. This requires that the conditions for the  
application of epenthesis be dependent on the potentially  
counterfactual application of assimilation.
In SP-OT, the derivation of a form is computed by generating and  
comparing multiple complete derivations (= output candidates) of that  
form in parallel, typically all but one of which are counter-factual  
(= nonoptimal). This fundamental property of SP-OT uniquely allows  
for the kind of 'looking forward' that is needed in order to properly  
express counterfactual derivation as just described. Although the  
relevant facts can be generated, counterfactual derivation itself is  
simply not statable in rule-ordering terms. The right result cannot  
be gotten with either of the two possible orders between epenthesis  
(stated so as to apply between identical consonants only) and  
assimilation.
It isn't clear precisely how the theory should be modified in order  
to correctly express this sort of rule interaction, but it must be  
something along the following lines: the derivation of a form must be  
allowed to pursue a counterfactual path, the end result of which may  
serve as the motivation for the application of a rule in the "real"  
path of the derivation that leads to the actual surface form. In  
other words, separate candidate forms must be derived and compared,  
much as they already are in SP-OT.