I’ve just finished a set of revisions to my paper previously entitled “Opacity deconstructed”, and now entitled “Opacity and ordering”. This new version is available in the same place as the earlier one, and the earlier one is still there as required by the lingBuzz archiving policies.
Note that lingBuzz also disallows editing of the title and abstract, but editing of the (eventual) publication venue is allowed. I edited that because this paper is no longer to appear in the Blackwell Companion to Phonology; it will instead appear in the 2nd edition of the Handbook of Phonological Theory. I’d like to thank the editors of both of these collections (Marc van Oostendorp, Beth Hume, and Jason Riggle in particular) again here, not only for their help with my own paper but also for their work on each collection as a whole. They both promise to be truly excellent and I know that everyone who reads this is as eager as I am to see them when they are finally published.
My paper is substantively and organizationally almost unchanged, but I’ve attempted to address very helpful comments and suggestions from several people who carefully read the earlier version (all acknowledged in the title footnote). The biggest change is that this new version engages in far, far less of the framework comparison that the earlier version engaged in, and instead focusses more on the main points of the paper: the typology of opacity and the decoupling of opacity and ordering. I hope readers of both versions will agree that the paper is much improved as a result.
At one point I was tempted to amplify the framework comparison angle by providing more detail on the “how it’s been done in OT” sections, but I eventually saw three problems with this. First, I think that the paper is long enough as-is; it would have been way, way too long with all this stuff added in. Second, I personally prefer papers — especially ones for collections like the one this will appear in — that don’t try to make too many points at once.
Third, I’ve come to the conclusion that engaging in framework comparison is in most respects a fool’s errand. One particularly frustrating reaction to my 2007 Phonology paper, for example, was the dismissive comment by Vaux (2008: 31) that I had made “a number of interesting modifications” to Kiparsky’s definition of opacity. I had done no such thing; what I did was show some of what I show in the present paper: that there is more to Kiparsky’s definition of opacity than counterfeeding and counterbleeding. No modifications were made because none were necessary. I’m not saying that Vaux deliberately misinterpreted or misrepresented my work, but I am saying that when it comes to framework comparison, there are some who will simply read into it what they want to read into it.
[ Another example of this sort of thing is summarized in this part of Opacity Wars discussion: there are folks who see e.g. the Elsewhere Condition as a some kind of natural outgrowth of rule-based serialism but e.g. Local Conjunction as stipulative “gymnastics” within OT (or vice-versa). I’m still completely baffled by this attitude; it would be funny if it weren’t so indefensibly asinine. ]
ANYWAY, I hope that the choices I’ve made in revising the paper will help to make its point more clearly to those who are receptive to it. One choice I made, for better or for worse, was to remove an extensive quotation and my commentary on it from the Vaux (2008) paper cited and linked above; just for kicks, an alternative conclusion with that quotation and commentary is copied below. Enjoy.
The resurgence of research on phonological opacity over the past fifteen years or so has unfortunately not paid attention to the substantive questions of learnability raised by Kiparsky’s original hypothesis; opacity has instead been wielded as a weapon in the larger debate between proponents of rule-based serialism and proponents of alternative theoretical frameworks, Optimality Theory in particular. The debate has been sharply polarized in most respects, but there is one mistaken ‘fact’ on which nearly all researchers on both sides mysteriously appear to have decided to agree: that rule-based serialism, via its central principle of rule ordering, uniquely offers a unified account of opacity as originally defined by Kiparsky. Consider, for example, these recent remarks on the issue by Vaux (2008: 38-39):
[T]he [rule-based serialism] treatment of opacity is significantly more elegant than its OT counterparts: it predicts exactly the attested types of opacity effects and deals with them straightforwardly and in a unified way […]. Since opacity is one of the most fundamental phenomena in human language, we must prefer a theory that accounts for it straightforwardly ([rule-based serialism]) over one that seems unable to deal with it (OT).
This passage, and the relevant section of the paper in which it is embedded, drips with unsubstantiated claims, both explicit and implicit: elegance and its significance, what is attested and what is not, what is opaque and what is not, what is straightforward and unified and what is not, what is fundamental and what is not. Vaux continues:
Some supporters of OT have responded that what [rule-based serialism] treats as a unified phenomenon, opacity, is actually a heterogeneous set of unrelated facts that are only made to look like a coherent whole by the theory. My response to this is that […] one fact needs one explanation. Our linguistic intuition, be we derivationalists or OT supporters, suggests that grammars involve generalizations that may conflict with one another; [rule-based serialism] provides a more successful account for this fact. One could add that, all else being equal, a theory that accounts for a range of phenomena via a single mechanism is to be preferred over a theory that accounts for the same facts with two or more mechanisms.
I demonstrate in this paper that it is not even the case that rule-based serialism “treats [opacity] as a unified phenomenon”, unless we decide to depart from Kiparsky’s agreed-upon definition of opacity and instead stipulatively (and perversely) define it as just those opaque interactions that can be described with rule ordering. Further discussions of the implications of opacity for theoretical framework comparison should either acknowledge this or provide a new, principled definition of opacity on which to base such discussions.