I only found out about this now. Thomas McFadden has posted his dissertation (University of Pennsylvania, April 2004): The position of morphological case in the derivation: A study on the syntax-morphology interface (you can also download chapters individually from his web page).

The main thesis of this dissertation is that morphological case is a purely morphological phenomenon, determined exclusively within the post-Spell-out portion of the derivation on the branch leading to PF. As such, case will depend in large part on the output of the pre-Spell-out narrow syntax, but the narrow syntax will not be able to make reference to or depend in any way on morphological case. I motivate this thesis by presenting extensive evidence that, contrary to what has been assumed since the late 1970s, morphological case is completely independent of the principles of positional DP-licensing that have been called syntactic or abstract Case. I then examine a series of syntactic phenomena which have been argued to depend crucially on morphological case. Specifically, I demonstrate that the interpretation and syntactic behavior of DPs marked with semantic and inherent cases is not due to their special case-marking. Rather, these DPs are distinguished from others by the syntactic structures in which they appear, and it is these structures that are responsible for both their special case-marking and their special syntactico-semantic behavior. I also present a series of empirical and theoretical arguments against making the syntactic processes that derive word-order freedom directly dependent within the synchronic grammar on rich morphological case-marking. I then develop a theory of morphological case-assignment, and show that the treatment of the actual morphology need make no reference to operations that are proper to the narrow syntax, either for the determination of which cases will appear or for the placement of the case-markers themselves. Finally, I reconsider syntactic Case in the light of the other results of the dissertation and explore the possibility that it could be eliminated from the theory, showing that, even for the regulation of subject positions in embedded clauses, where it is supposed to play its most important role, it offers no real insight into the distribution of DPs. At best it is largely stipulative, and at worst it makes the wrong empirical predictions.