This is a follow up to a quick comment I left in the Reading Group thread. I am not entirely up on the history of the field, so maybe these points are trivial. If so, excuse me.
I found the discussion of rule ordering in section 5 to be interesting. There seem to be a couple of issues that popped up with regard to rule ordering in the 1940s. One is historicity–how seriously are we going to take the time/motion metaphor? Another is the issue of primacy–if a, b, and c are derivable from one source, which one, if any, is primary? And a third is Harris’ claim that extrinsic rule ordering masks natural relationships between classes of derivations.
The first and last issues seem especially interesting after the Mr. Verb Kerfluffle. One of the things that was suggested there was that if you have rules, rule ordering is natural. Goldsmith shows that for some phonologists in the 1940s, rule ordering wasn’t a natural step at all. And it seems to me that a lot of phonology after SPE was concerned with addressing that last bit–making the rule ordering natural (there might be something about the Elsewhere Condition here, but I don’t feel qualified to talk about it).
What put me in mind of the richness of the base (RoB) was the middle part about primacy. RoB is the OT claim that the set of possible inputs to the grammar is universal, thus getting rid of the issue of primacy. In the hypothetical case of a, b, and c the grammar has to make sure that whatever the input /a/, /b/, /c/, etc., nothing maps to b in an environment where b is disallowed. Although RoB doesn’t rule out the use of archiphonemes (or underspecification) it does make them seem unneccesary since you can construct a grammar that will always map a and b to c in the appropriate context for example.
Ed’s half-right here, I think: RotB (or RoB, wevs) certain changes the interpretation of primacy, but primacy is still an important element of the analysis of any process in OT.
In the case of an alternation-based neutralization process (e.g. [-son] → [-voi] / __#), for instance, you still have the alternation evidence (alternating voiced-voiceless pairs and nonalternating voiceless pairs) to show you that voiced obstruents are basic (= primary) to the alternation. In the case of complementary distribution, one of the allophones can still (and in many cases must still) be analyzed as basic and the other(s) derived.
But the upshot that Ed notes is completely correct: under the RotB hypothesis, primacy is not encoded in the input to the grammar, but emerges as a property of constraint interaction. OK, so I guess he’s more than half-right. Sorry Ed.
Wasn’t primacy only an issue when there was no alternation evidence?
Kerfluffle? Is that an example of aggressive reduplication?