Current Issues in Linguistic Theory 114 ) THE EPSILON OPERATOR AND E - TYPE PRONOUNS

0. Introduction In this paper we attempt to provide a homogeneous description of definite and indefinite noun phrases. We therefore assume one semantic representation for one interpretation that the definite article the and the indefinite article a share. Our starting point is the classical theory of descriptions as developed by Russell in his famous article of 1905, " On Denoting. " In the first section we introduce the basic assumptions or intuitions that form the foundation for our considerations. Russell's analysis will be shown to be too restrictive to deal with natural language phenomena. The exposition of the intuitions will be more comprehensive than is usually the case, as we think that for a formalism devised for natural language description our basic ideas about language are at least as important as correctness and elegance in the formal model. Furthermore, it will become apparent that some other approaches to the descriptive content of common noun phrases actually fail because they do not pay sufficient attention to basic characteristics of natural language. In the second section a choice function, namely, the epsilon operator suggested by Hilbert, will be introduced and modified to capture the content of definite and indefinite descriptions adequately. Its properties will be compared with the basic ideas set forth in section 1. This will include a closer look at the well-ordering of sets and the dependency of descriptions on situations. In addition, we will introduce simple rules of thematization and rhematization that allow one to specify the description or to extract a predicate from a description. These rules are largely guided by the actual use of natural language. It will become clear that the modified epsilon operator is much more appropriate to the variations of actual language use than Russell's approach with the iota operator. In the third section we apply the analysis proposed in the second section with the help of epsilon to E-type pronouns. We will show that our analysis solves the problematic bishop-sentences, since it does not imply any uniqueness condition. The ideas developed in this paper go back to Egli (1991) (in circulation since 1989).