The purpose of this paper is to build a model that gives a natural explanation of the distribution of demonstratives in the anaphoric use. Discourse Management Theory (DMT; Takubo and Kinsui (2000) among others) provides a persuasive model for this issue. However, this theory faces some difficulties, especially when dealing with sentences such as those presented in Iori (1997) and Tsutsumi (1998). I argue that this is because the theory is strongly based on the concept of ‘experience, ’ however, these problems are solved when we adopt the model presented in this paper.
We create (at least) two worlds in our mind: one of these is Ws and the other is Wp. Ws is the world in which we refer to things directly (such as proper nouns and things that the speaker believes to exist in the real world (Wo)) while Wp is the world where every element is altered to a variable (x, y). Following Tsutsumi (1998), we assume that Ko/A-series refer to elements in Ws and So-series to those in Wp.This predicts that if an object is registered in both of the worlds, then both Ko (A) /So-series can be used. If it is registered in only Ws or Wp, then only Ko (A) or So can be used respectively. This idea is reminiscent of that of DMT's, however, by abandoning the concept of ‘experience, ’ and introducing the idea of ‘variable’ for the So-series demonstratives, we succeed in explaining a wider range of data than DMT can deal with.
