I appreciate the critical analysis that Adam Ussishkin and Natasha Warner make of my posting, A Leap of Faith? Their proposed typology of research questions is an explicit and detailed follow-up that clarifies many issues that my original posting had only left implicit. Regarding the questionable relationship among Steps 3, 4, and 5, I believe that I had already acknowledged, in response to ACW’s initial comment, that to make such a leap is indeed an unwarranted oversimplification.
Eric’s interpretation of the last paragraph of my posting is correct. By using the idiomatic expression “…and call it a day,” I was trying to give an almost tongue-in-cheek flavor to the preceding clause, “that one could put everything in the lexicon…” The conversational implicature that I wanted to generate is exactly the one that Eric inferred, i.e., “not forming any generalizations, period.” Of course I am well aware of other representational possibilities, e.g., exemplar-based models that call for the inclusion of predictable and rich phonetic detail in lexical representations while still allowing generalizations to be formed over this information.
Following up on Eric’s discussion of “hybrid” approaches within the job market and the field, I’d also like to point out that the marriage of formal models and phonetically-based, experimentally-motivated models sometimes makes it possible to gain new insights into how one relates to the other. Browman and Goldstein’s Articulatory Phonology posits articulatory gestures as units of articulation as well as phonological primitives. Some people assume intrasegmental (and some, even intersegmental) timing to be present in the gestural scores of lexical items. But as John McCarthy points out in footnote 11 of Comparative Markedness, such an assumption runs counter to Richness of the Base. And as Nancy Hall argues in Chapter 1 of her dissertation, gestural OT analyses that assume faithfulness to lexically-specified timing overpredict universally unattested contrasts based on the presence vs. absence of, e.g., intrusive vowels and consonant release. Her conclusion is that UG must not contain such a faithfulness constraint (although I’m sure some people might make the counterargument that this is just stipulation). So while nothing prevents language users from storing predictable intersegmental timing in the lexicon, the nature of the constraint set ensures that such information will never be phonologically contrastive per se. It seems to me that this implication about CON could not even have been made in the absence of “hybrid” attempts to combine OT and Articulatory Phonology.