2. Hoji 1985
　　The papers in this volume were published during 1995-2003. Papers 1-6 draw heavily from Hoji 1990, which is in turn a continuation of Hoji 1985. In retrospect, Hoji 1985 tried to identify the informant intuitions that are necessarily based on the satisfaction of a c-command condition. I was concerned mainly with the (un)availability of bound variable construal and scope dependency in Japanese that seem to be sensitive to (i.e., seem to require the satisfaction of) a c-command condition. By making reference to the (un)availability of the dependency interpretations in question, I argued for a particular view of the phrase structure of Japanese that it is strictly binary-branching.
　　My main concern in Hoji 1985 was to express/describe some "phenomena" in Japanese in the terms of the theory I adopted at the time and to consider what I might possibly be able to say about the theory on the basis of my "findings" in Japanese. The empirical generalizations put forth (or adopted) in Hoji 1985 are, however, often far from being robust.
　　It seems to me to be reasonable to say that my research subsequent to Hoji 1985 started out as an attempt to overcome the shortcoming of Hoji 1985 that the empirical generalizations put forth (or adopted) there are often far from being robust, despite the fact that they have been accepted in much of the subsequent generative research as one of the basic sets of generalizations in Japanese, along with the proposed/assumed structural analyses for the sentence patterns in question.
　　Focusing on BVA, it has become clear to me over the years that we could replicate the robust informant judgments as schematized in (1) (and other related paradigms) only if we used certain types of expressions for A and B in BVA(A, B).

(1)a.A-ga ... [ ... B ... ]-o ... V-T
with BVA(A, B)
b.*[ ... B ... ]-ga ... A-o ... V-T
with BVA(A, B)
c.[ ... B ... ]-o ... A-ga ... V-T
with BVA(A, B)
d.A-o ... [ ... B ... ]-ga ... V-T
with BVA(A, B)
e.[ ... B ... ]-ga ... A-o ... V-T
B is referential.

The empirical thesis I pursued over the years can be summarized roughly as follows: the use of the "right types of expressions" for A and B of BVA(A, B) makes it possible to obtain robust informant judgments in line with the patterns indicated in (1) and that we obtain not-very-robust informant judgments when "wrong types of expressions" are used as A and/or B of BVA(A, B). Ueyama 1998 proposes what this means, on the basis of her theory of anaphoric relations and her proposal about how the OSV order in Japanese is derived.
　　In Hoji 1985, I tried to establish paradigms instantiating the generalization in (1), and a similar generalization for DR, with a number of different expressions for A of BVA(A, B) and also for A of DR(A, B). All the papers contained in this volume are concerned with the issue of how we can establish the robust empirical generalizations under discussion, directly (when they deal with BVA(A, B)) or indirectly (when they deal with the sloppy-identity reading).
　　The evidence presented in Hoji 1985 for the binary-branching thesis for Japanese was based on the distribution of BVA, DR, and coreference. The distribution in question was identified in the terms of, i.e., by adopting, theories that make reference solely to c-command. At that time, there had been theories of anaphoric and/or scope dependency based on precedence (combined with some structural relation, such as c-command). If the relevance of precedence had been accepted (along with the relevance of c-command) in the description of the distribution of BVA, DR, and coreference, it would not have been possible to construct arguments for the binary-branching thesis for Japanese in Hoji 1985 on the basis of the distribution of BVA, DR, and coreference. It is in this sense that the arguments in Hoji 1985 for the binary-branching thesis for Japanese were circular, as pointed out by Fritz Newmeyer (p.c. 1985).
　　Reinhart 1983: Chapter 7 suggests that what formally underlies the BVA(A, B) also underlies the sloppy-identity reading and that the availability of BVA(A, B) and that of the sloppy-identity reading are both constrained by the local disjointness condition, widely known (then) as Principle B of the Binding Theory, as well as by the c-command condition. The discussion of the paradigms of the sloppy-identity reading in English in Reinhart 1983: Chapter 7, among other works, prompted me to see whether we could clearly observe the effects of the structural conditions, i.e., the c-command condition and the local disjointness condition, on the availability of BVA and the sloppy-identity reading in Japanese. In Hoji 1990, I continued my attempt in Hoji 1985 to establish the crucial relevance of c-command for BVA(A, B) while at the same time searching for the best choices of A and B for BVA(A, B) and the best choices of the relevant expressions in the sloppy-identity reading context for the purpose of obtaining as robust judgments as possible, as a reflection of formal properties of the language faculty. My concern was how to demonstrate that BVA(A, B) and the sloppy-identity reading (of the "right type"―see above) are regulated by the same structural and lexical conditions. Such considerations in turn lead us to consider the properties of various sentence patterns and give us ideas about how those sentence patterns should be formally represented.