Fig IV

Pegs: The ultimate epicycles of dynamic logics? Some comments on "Coreference and Modality in the Context of Multi-Speaker Discourse" Klaus von Heusinger 1997 In: H. Kamp & B. Partee (eds.). Proceedings of the Workshop "Context Dependence in the Analysis of Linguistic Meaning". Prague and Stuttgart 1995.

• Contents
• Introduction
• PDF-File (65K)

Contents

1. Introduction
2. The Platonic Cycle: Anaphora as Binding
3. The First Epicycle for Extending the Scope: Dynamic Interpretation
4. The Second Epicycle for the Referential Challenge: Pegs
5. The Third Epicycle for Modal Subordination: Accommodation
6. The forth Epicycle for Multi-Speaker Discourse
7. A different Picture: Elliptical Constructions

Introduction

In their paper "Coreference and Modality in the Context of Multi-Speaker Discourse", Groenendijk & Stokhof & Veltman (= G&S&V) (1996b) extend their analysis of anaphoric relations across sentence boundaries to multi-speaker discourses. They sketch a formal account of the use of anaphoric pronouns across different speakers. This very ambitious enterprise opens an additional field of data and provokes new linguistic discussion. They apply an enriched version of their classical dynamic logic in order to cover the new ground. The classical dynamic logic (Groenendijk & Stokhof 1991) represents the meaning of sentences as the potential to change the information of the speaker and hearer. It is formally represented as the relation between information states or assignments. The enriched version (G&S&V 1996a) extends this formalism by a so-called referent system that introduces "pegs" as intermediate discourse entities. Pegs are connected with the variables in the object language, on the one hand, and with the objects in the model (of the real world), on the other hand. For the descriptions of a multi-speaker discourse, G&S&V (1996b) introduce the notion of an utterance by a speaker. Examples (1) - (3) (their (18) - (20) in section 3.3) illustrate the increasing complexity. In (1) the sequence of the two sentences is dealt with as a conjunction. In (2), it is dealt with by the sequence of two utterances by the same speaker, and in (3) by two utterances of different speakers.

 (1) A: A man is walking in the park, and he is wearing blue suede shows.
           <a, $x Px & Qx>
 (2) A: A man is walking in the park.
      A: He is wearing blue suede shows.
          <a, $x Px>, < a, Qx>
 (3) A: A man is walking in the park.
      B: He is wearing blue suede shows.
      <a, $x Px>, <b, Qx>

In order to license anaphoric relations across different speakers, as in (3), they make additional assumptions as to the possibility of introducing pegs by one speaker and referring back to them by another speaker.
 In this comment, I argue that this additional assumptions are not forced by the new data, but rather by the misconception of the classical dynamic theory in Groenendijk & Stokhof (1991). I am not going to criticise dynamic logic entirely, but one essential aspect of dynamic logic used in G&S&V (1996a; 1996b), namely the treatment of anaphoric pronouns as bound variables. This neat and simple picture is very tempting to logicians. Simple cases of anaphoric relations can be explained by this approach. However, more complex linguistic data exhibit mismatches between the linguistic phenomena and this "platonic idea of anaphora". I show that even a quite sophisticated bound variable analysis cannot cover all the relevant linguistic data. The dynamic derivative of the bound variable approach does not really explain the real character of anaphora but rather hides it by introducing additional complex constructions instead of changing the basic picture.
 This is not a criticism of the dynamic enterprise as such, which is certainly a very powerful tool for analysing natural language meaning. However, I argue that the use of bound variables is not only not necessary, it even hides the deeper structure of anaphora. G&S&V (1996c) reacted to a similar criticism expressed in Peregrin & von Heusinger (1996) in a very moderate way. They granted that they do not see any need for syntactic variables, but they still assume that semantic variables (pegs, discourse referents) are necessary: "... contra Peregrin and von Heusinger, we hope to have shown that giving up syntactic variables should not involve throwing away semantic variables" (G&S&V 1996c, p.23, section 12.1). Even if this would be their definite position (which is not so obvious from other papers) my argument would concern the following fact: Since they stick to the intuitions behind the classical binding approach they only respell the classical theory - with or without syntactic variables. Hence, they cannot free themselves from the limitations of the binding view. And this is the topic of my comment.
I illustrate this epistemological process of sticking to an idea and trying to maintain it even against increasing evidence, by quoting from the crisis in astronomy at the beginning of modern times.  Back then, people stuck to the very Platonic idea and the Aristotelian dictum that any movement must be idle. The movements of the planets had been understood only in terms of circular orbits. Observations that differed from this abstract view were explained by introducing a more complex system consisting of different levels of epicycles. Nevertheless, no number of epicycles was able to explain all "irregular" movements that were gained by more advanced observations. Eventually, Kepler (1571-1639), who was born in Weil der Stadt near Stuttgart and who later started his career as a student of theology at the (Lutheran) University of Tübingen, proposed his famous three laws of planetary motion that describe the planet orbits as ellipses instead of circles. "They put an end to the Aristotelian dogma of uniform motion in perfect circle, which had bedeviled cosmology for two millennia" (Koestler 1967, 330). They formed the fundament of all modern theories of planet movement.
 Analogously, I argue that the assumption that anaphora has to be represented as binding is logically very simple. However, it rather hides the linguistic nature of anaphora. No sophisticated modification of this idea can explain the variety of anaphoric phenomena. Only a different view can give a deeper understanding of anaphora. In this alternative, anaphoric pronouns are not represented as bound variable, but rather as terms.

 In section 2, I give a rough characterisation of the fundamental problem concerning anaphora and sketch two basic approaches. I then try to follow the development of dynamic logic and its intricacies by the example of anaphoric pronouns in multi-speaker discourse. First, I discuss shortly the idea of dynamic interpretation in section 3, and then I summarise and comment on the introduction of the new concept of "pegs" in section 4 as restoring the notion of referentiallity in a different manner. Thirdly, I analyse the treatment of anaphoric pronouns in modal contexts and show that even a sophisticated binding approach cannot cover the linguistic data in section 5. Finally, I sketch the treatment of multi-speaker discourse in section 6 and show that the problems we have then already encountered reappear in this complex analyses. In the last section I tend towards a different approach that goes back to Evans' critic against Geach and that is a more elliptical analysis.

PDF-File (65K)

Zurück an den Beginn der Seite