Opacity

A couple of days back, someone commented on the post about Zombie Feynman, where I mentioned that folks have compared Optimality Theory in phonology with String Theory in physics. This led me to mention opacity, bringing this question:

How pervasive is opacity really? (I'm not trying to be obnoxious, I really want to know!) I'm still a newbie (i.e. lowly grad student), but every time I hear someone talk about opacity, they mention the same Hebrew examples.

First, if your question doesn’t start with ‘yo, asshole’ or ‘as every schoolboy knows’, I figure it doesn’t count as obnoxious. (And we live by the energy of grad students precisely because they almost always really want to know answers.)

OK, starting with some background for non-linguists. Opacity is a name for patterns that are not 'surface true'. Following Kiparsky's formulation, take this kind of process:

A → B/C__D

That is, some sound 'A' becomes 'B' when it appears between 'C' and 'D'. It's opaque if you do get A in this environment or if you get B for some other reason. This was long captured by using ordered rules. (McCarthy, in his Thematic Guide, aptly says that serial derivation allows us to "state temporary truths".) Here are two examples from Greg Iverson's "Rule Ordering", Handbook of Phonological Theory:

First, Canadian Raising for one variety of Canadian English (following Joos 1952) interacts with flapping in a crucial way:

• Raising: /ay, aw/ centralize to [əy, əw] before voiceless consonants (not voiced).
• Flapping: /t/ merges with /d/ between vowels to what we can characterize as [d].

If the rules apply in that order, you get rider [raydɚ] but writer [rəydɚ]. Looking only at the surface form, writer looks like it doesn't have the environment for raising but it does. (In some kind of Canadian English, it works the other way.) The 'temporary truth' would be the moment where writer has a /t/ to trigger raising, even though we don't see the /t/ on the surface.

Second, in Icelandic, the word for 'package' is böggli, from /bagg+ul+i/. The /u/ triggers umlaut (/a/ becomes [ö]) but it is then deleted, leaving its 'trace' on the final form. Here, it looks like you don't have the environment for umlaut, but you do.

And opacity is pervasive in human language. Derivations and Constraints in Phonology, ed. by Iggy Roca explores a set of examples from Dutch, Berber, Polish, Yokuts and so on, in addition to the now-famous Hebrew spirantization examples Idsardi treats there.

That answers the reader's direct question, but consider why this matters: In any monostratal theory (one without stages of derivation), getting these interactions is a huge problem. This isn't the place to run through them, but some readers will be familiar with sympathy theory, comparative markedness, and so on. I heard one person sum it up this way a few years ago:

Opacity is ubiquitous in human language, and earlier theories of phonology could deal with it easily. It's hard to see why those advantages have been abandoned for an approach that can't handle opacity without lots of gymnastics, if at all, for benefits that don't look all that great.

Those gymnastics remind some people of String Theory, I think.

Finally, I beg forgiveness for polluting the consistently low level of discussion here with a moment of theoretical linguistic seriousness, but I promise we'll return to the gutter faster than a 16-pound ball hurled by a hyperactive, sugar-high kid at a bowling alley birthday party.

Image from here.