Without an Index: a lexicalist account of binding theory

Binding principles delimit the relative positioning of anaphors and their admissible antecedents in grammatical structure. These principles stem from quite cogent empirical generalizations and exhibit a universal character, given the hypothesis of their parameterized validity across natural languages. Their mutual relations involve non-trivial cross symmetry, lending them a modular shape and providing further strength to the plausibility of their universal nature. In contrast to this, the encoding of anaphoric binding constraints into formal grammars has presented considerable difficulties. As an example of such difficulties, it is worth noting that the mainstream approach for this encoding, which dates back to (Chomsky, 1980) and is based on the exhaustive and overgenerating indexation of grammatical representations, has been shown to require extra-grammatical processing steps of non-tractable computational complexity (Correa, 1988; Fong, 1990). As for HPSG, we would like to obtain a fully-fledged integration of binding theory into formal grammars. In the nine page Appendix of (Pollard and Sag, 1994), the fragment of grammar developed and discussed along this book is formally specified using the HPSG description formalism: Binding principles receive a definition in Chapter 6, but are a major part of grammatical knowledge discussed in this book that escapes such formal encoding. While pointing out the fact that these constraints are waiting to be accommodated into HPSG grammars, Backofen et al. (1996, p.65) and Bredenkamp’s (1996) discussion of this issue implies that some kind of essential limitation of the description formalism for representing grammatical knowledge might have been reached. In particular, Bredenkamp presents a detailed discussion of some apparent options to encode binding constraints in HPSG — coindexing as