Fundamental issues in the theory of binding
Within the current framework of the Extended Standard Theory of generative grammar, the Government-Binding theory (henceforth GB) of Chomsky 1981, two lexical NPs are interpreted as having the same intended reference if they bear the same index. Under an optimally simple formulation, coindexing is achieved by a rule “index NP.” Thus there is no special rule of coindexing or any special prohibition against coindexing by the basic indexing rule (see Freidin and Lasnik 1981 for discussion). This rule generates representations of sentences where lexical NPs are interpreted as coreferential as in (1–3).
It is possible to generate different grammatical sentences by eliminating the coindexing in (1) and (2), but the elimination of coindexing in (3) yields an illformed representation. The expression each other in (3) is a lexical bound anaphor which has no intrinsic reference of its own and must get its reference from some NP antecedent in the sentence in which it occurs. The pronoun she in (2) also has no intrinsic reference; but unlike the bound anaphor, a pronoun can get its reference outside the context of the sentence in which it occurs. The proper name John in (1) presumably has an intrinsic reference—in which case coindexing in (1) does not fix the reference of the second occurrence of John, but rather establishes that both names designate the same individual. In what follows, bound anaphors and pronouns will be referred to as anaphoric elements and the NP with which each is coindexed in a sentence will be designated its antecedent.
The rule of indexing as formulated generates representations of anaphoric relations which do not occur—e.g., (4–6).
The examples in (1–6) show that coindexed NPs have a restricted distribution independent of their linear order (since this order is identical for each anaphoric element/antecedent pair). Moreover, the distribution of coindexed NPs differs according to NP-type. Coindexed names do not occur in the same contexts as coindexed pronouns (compare (4) and (2)), and coindexed pronouns do not occur in the same contexts as bound anaphors (compare (5) and (3)).
Within GB the limited distribution of coindexed NPs is accounted for by three general principles—one for each NP-type mentioned above. These principles, commonly referred to as Principles A, B, and C, are given schematically in (7).
These principles constitute a theory of binding. They are interpreted as conditions on syntactic representations at some particular level or levels in a derivation, rather than as conditions on the application of the indexing rule (i.e. on the indexing derivation itself). Under this interpretation the principles in (7) define wellformedness of binding representations.
An explication of the binding theory as instantiated in Principles A-C must address the following fundamental issues concerning the formulation and application of these principles.
The following discussion is meant to be illustrative rather than definitive. It is an attempt to elucidate current work on the binding theory from the perspective of the fundamental issues raised.
The remainder of this paper is organized as follows. Questions (i) and (ii) which deal with the structural relation between anaphoric elements and their antecedents are discussed in section 1 in terms of a single hierarchical relation, c-command. Section 2 covers the analysis of domain D of Principle A. The sproper characterization of domain D′ of Principle B as differentiated from domain D of Principle A is the focus of section 3. Section 4 deals with the application of binding theory to empty categories and the related issue of the typology of empty categories. Section 5 contains a preliminary discussion of question (vi). The determination of what levels of representation the various binding principles apply to turns out to be a complicated matter. The final section provides a brief summary of the major points in the preceding sections.
