Thanks to all for the discussion and critiques concerning chapter 3 of my in progress ms. While I had hoped that this particular ms. would be able to avoid all specifics when it came to morphological theory, I am once again convinced that it is impossible to say anything clearly without being specific about one’s implementation.
In particular, I have been forced to think about how to implement what I need in a system that respects separation and is in general consistent with a late insertion model. In addition, I want it to be a spanning system a la Bye and Svenonius 2011, Adger 2009 etc. and not a DM-like architecture. As we discussed, I had also set myself the task of unifying the participle qua lexical item across all of its uses, but at the very least the different `passive’ uses.
Antonio’s solution to my EN/ED problem in class was to say that EN/ED was the spell out of a verb that had underassociated features in the sense of Ramchand 2008. I liked this idea because it seemed to conform best to the intuition I was trying to capture. However, when I tried to make it work I kept getting myself back into lexicalist paradoxes. In the end, I think I have come up with a better version which I describe briefly here. First of all, I claim that the participle is indeed a lexicalization of a contiguous span of heads in the Event Essences domain, but we do not need to inspect the verbal entry to know when it kicks in. I stipulate that normal verbal entries in addition to their init, proc, res features also have Evt and Asp. Bare uninflected roots are the same but without Asp. Neither of these two can freely underassociate. The participle in EN/ED is a span lacking Asp, but any span lacking Asp, so that it does in fact freely underassociate within those limits.
I think I can still have my Poser blocking for passives of unaccusatives with this implementation.
In the new reworking, the eventive participle does in fact spell out the whole initP, but I have separated CAUSE from the introduction of the External argument, which now happens in Evt (in conformity with most of the literature on argument structure since Pylkkaenen 99, but in opposition to what was assumed in Ramchand 2008). This helps me out of the event entailments problem pointed out by Serge. It means that when the eventive passive is instantiated at Asp there will in fact be a causing event that gets instantiated, just not the external argument. I tried to think about whether that made any predictions, given that causer and causing event have now been split in this way but I got stuck. Anyone who has any ideas please tell me. But this at least makes the problem raised by Serge much better.
I now have a completely rewritten version of the first half of chapter 3 for anyone who is interested in seeing it. There will be implications for the perfect, but I will rewrite that this weekend.