So what does all this mean? My interpretation of the situation is as follows.
First, I think that the generalization captured by my analysis of these extremely well-known (to the point of being almost mundane) English facts is a significant one that has been completely missed in previous analyses: Epenthesis ignores [±voice] not arbitrarily, but because [±voice] assimilation would be independently expected to apply if Epenthesis didn’t. This is line with much recent work (e.g. Côté 2004, Frisch et al. 2004) on the nonarbitrariness of segmental similarity and stands in stark contrast with the opposing view, most recently and explicitly articulated by Reiss (2003), that ‘sufficient identity’ is defined arbitrarily for every rule.
Second, the only way to express the relevant generalization is through opaque process interaction of a particular type that I believe has never been considered before, and that I am claiming here is only possible to express in derivational phonology with significant (and arguably ad hoc) modifications to the theory. In OT, as I’ve shown above, opacity of this type can be analyzed with no particular difficulty, no modification to the standard theory, etc.
In what way is this process interaction opaque? Kiparsky (1973:79), building on Kiparsky (1971), defines opacity this way (see also McCarthy 1999:358, from which this concise statement is adapted):
A process P of the form A → B / C __ D is opaque if there are surface structures with any of the following characteristics:
- instances of A in the environment C __ D.
- instances of B derived by P that occur in environments other than C __ D.
- instances of B not derived by P that occur in the environment C __ D.
The opaque process P in this case is the epenthesis process, which in the analysis I advocate would not be stated as repeated here:
- ø → V / C1 __ C2 #,
where C1 and C2 differ at most in their value of [±voice].
Rather, it would be stated this way:
- ø → V / C1 __ C2 #,
where C1 = C2
The way this process is formulated here, it is opaque according to the second of Kiparsky’s characteristics: B derived by P is the epenthetic vowel, and the environment other than C __ D in which it appears is t __ d #.
In McCarthy’s terminology, this is a case of non-surface-apparent opacity, the kind of opacity that is typical of counterbleeding rule order. Ironically, McCarthy (1999:332) states that “[u]nless further refinements [to standard OT] are introduced […] OT cannot contend successfully with any non-surface-apparent generalisations”. (See note.) As I think I’ve shown here, it may be true that OT can’t deal with counterbleeding but it doesn’t seem to be the case that OT can’t deal with all cases of non-surface-apparent opacity. In fact, this is a turn-around on the usual play: what I’ve shown is that there is a type of opacity that standard OT can deal with very successfully while derivational phonology cannot (unless, of course, “further refinements are introduced”).
A final remark is in order here. Notice that the opacity of the epenthesis process (viz. its application in the context t __ d #) contributes to the greater transparency of the generalization responsible for devoicing: if epenthesis didn’t apply, and if devoicing could not create adjacent identical consonants, then the output would be the voice-disagreeing td#. Thus there is a kind of conspiracy (Kisseberth 1970) going on: both devoicing and epenthesis apply (the latter only in part) to avoid voice-disagreeing final clusters. This challenges Kiparsky’s (1973:80-81) strong conclusion:
The explanation of conspiracies is thereby reduced to the theory of opacity. The fact that
(2-13) Languages tend to have conspiracies.
follows from the more general fact that
(2-14) Languages tend to have transparent rules.
On the other hand, the analysis under discussion here appears to be a case of the type of conspiracy that Kiparsky (1973:78) describes this way:
[A] phonological rule can function as part of a conspiracy indirectly, by causing or preventing the application of other rules in conformity with the target.
I’m not aware of any other examples that can be described this way; unfortunately, Kiparsky doesn’t specifically cite any. As usual: if you know of any, please let me know.
Comments most welcome!
Côté, Marie-Hélène. 2004. Syntagmatic distinctness in consonant deletion. Phonology 21. 1-41.
Frisch, Stefan, Michael Broe & Janet Pierrehumbert. 2004. Similarity avoidance and the OCP. NLLT 22, 179-228.
Hill, Jane. 1970. A peeking rule in Cupeño. LI 1, 534-539.
Kiparsky, Paul. 1971. Historical linguistics. In W. Dingwall (ed.), A Survey of Linguistic Science, 576-642. College Park: University of Maryland Linguistics Program.
Kiparsky, Paul. 1973. Abstractness, opacity, and global rules (Part 2 of “Phonological Representations”). In O. Fujimura (ed.), Three Dimensions of Linguistic Theory, 57-86. Tokyo: TEC Company, Ltd.
Kisseberth, Charles. 1970. On the functional unity of phonological rules. LI 1, 291-306.
McCarthy, John. 1999. Sympathy and phonological opacity. Phonology 16, 331-399.
Reiss, Charles. 2003. Quantification in structural descriptions: Attested and unattested patterns. The Linguistic Review 20 (Special issue: Typology in Phonology), 305-338.
This quote appears in the following context of McCarthy (1999:332), emphasis added:
As OT is currently understood, though, constraint ranking and violation cannot explain all instances of opacity. Unless further refinements are introduced, OT cannot contend successfully with any non-surface-apparent generalisations nor with a residue of non-surface-true generalisations.
Most phonologists are well-aware that standard OT generally has difficulty with opaque process interaction, but the emphasized part of this quotation is frequently overlooked or misunderstood. There are certain non-surface-true generalizations that OT handles quite simply; see e.g. McCarthy’s (1999:363ff) discussion of non-surface-true opacity of the chain shift (“counterfeeding on focus”) type, which is not problematic, as opposed to non-surface-true opacity of the “counterfeeding on environment” type, which is problematic and requires significant (and arguably ad hoc) modifications to the theory. (Back to text.)