Accommodating Topics

This paper concerns the relevance of notions of sentence topic and discourse topic to the analysis of sentences containing presuppositions. Firstly I consider sentences where quantificational determiners quantify-in to presuppositions. By considering texts containing such sentences, I show that intermediate accommodation cannot be triggered by presuppositions, contrary to the predictions of van der Sandt’s recent model. However, a process I refer to as topical accommodation could justify the existence of the readings predicted by van der Sandt’s model in some cases. I then show that similar problems occur in the treatment of presuppositions occurring in the consequents of conditionals, and once again conclude that current models err by not taking into account topic-focus articulation and issues of discourse coherency. 1 The Naive Informant How is a naive informant to guess what the relevant topic of conversation is when presented with a decontextualised single sentence example? The mysteriousness of the way in which people “make up a context” for such examples is generally recognised to be problematic for standard linguistic methodology. When we come to studying aspects of meaning which specifically concern the previous context — I would call all such aspects of meaning presuppositional — the problem becomes acute. The question of what a sentence presupposes becomes a question of what propositions hold in normal contexts of utterance of the sentence. But what does normal mean? The standard tests for presupposition are doubly problematic in this respect. Although the informant is asked about implications and not presuppositions, two sentences rather than one are involved (eg. a sentence and its negation). That the informant is asked about implications cannot disguise the fact that they are presuppositionally derived, that what we are really after is propositions that may be presumed to hold in normal contexts of utterance. Only now we are concerned with normal contexts of utterance of two sentences, and not one. Again, what is a normal context of utterance? Exactly what ceteris have to be paribus across the contexts of utterance for the two example sentences? I will suggest that it is vitally important to consider explicitly the discourse contexts in which presuppositional example sentences occur. 2 Presupposition and Quantification I now want to draw attention to one particular aspect of the elegant theory of presupposition presented in [van der Sandt 92], namely the way in which presuppositions can trigger quantificational domain restriction, or something like it. I will argue that this aspect of the theory is not sustainable once the relevance of discourse context to the interpretation of example sentences is taken into account. Consider the following example: E1 Every German woman drives to work in her car. According to conventional wisdom, the NP her car carries a presupposition of car ownership. Exactly what happens to this presupposition in a case like E1, when, on the reading I am interested in, it is bound by a quantifier, is a moot point. Van der Sandt proposes that in order for the sentence to be understood, the presupposition must be accommodated, in something like the sense of [Lewis 79]. People’s tastes in accommodation vary. Since van der Sandt houses presuppositions in boxes designed by Kamp, the well known Dutch representationalist, the process of accommodating a presupposed proposition becomes a matter of moving the representations of presuppositions from box to box. A suitable box must be found somewhere along the accessibility path leading up to the presuppositional expression. The notion of accessibility is just that of [Kamp 81], whilst the details of exactly what is a suitable box for a presupposition has been discussed by van der Sandt at length, mostly in terms of his correctness conditions on DRS’s. I will not comment on these correctness conditions, but merely note one requirement van der Sandt places on accommodation: a proposition free in some discourse marker can only be accommodated in a location from where the marker is already accessible. Rather than considering accommodation in the abstract, let us look at a concrete example: the boxes that can be built to represent E1. After resolution of the possessive pronoun her, something like the following DRS is built, where a presupposed DRS is distinguished by its extra thick walls: