Buoyancy and Strength

The chief characteristic of presuppositions is that they tend to take wide scope, yet most theories of presupposition, the author's not excepted, fail to provide an explanation of this fact. Recently, however, it has been suggested that a principled explanation can be given in terms of informativeness: the idea is that presuppositions simply prefer stronger readings to weaker ones. This proposal is studied in some depth, and is shown to lack solid empirical evidence. Furthermore, it is argued that assuming a preference for strong readings is either ad hoc, when restricted to presuppositions, or just false, when held to apply more widely. The paper leaves the main problem very much where it is, though some suggestions are made as to how the situation might be improved. Anaphoric pronouns may be used to refer back to an object introduced in the preceding discourse, but also to refer forward to an object yet to be introduced (the latter usage is sometimes called 'kataphora' or even 'cataphora'). This may be so, but a theory of anaphora that predicts only this much is seriously incomplete, because it does not account for the fact that forward reference is the exception and backward reference the rule. A theory that does not explain this preference is not much of a theory at all. Most of the phenomena a theory of interpretation has to deal with are like this. It would be very nice if we had theory that predicted all and only possible interpretations for any expression in any context; but it would hardly count as a full-fledged theory of interpretation. Small wonder, therefore, that the framework of optimality theory naturally suggests itself for dealing with a wide range of problems in semantics and pragmatics. One of the problems that immediately comes to mind is that of presupposition projection, not only because projection phenomena seem to spring from the interaction between several forces of varying strength, but also because that is precisely how many theories of presupposition treat their subject matter. Optimality-theoretic treatments of presupposition have been proposed by Zeevat (1999) and Blutner (this volume), and in the following I will assume that some account along these lines is the right one. What I will be worrying about is one of the constraints postulated by Blutner and Zeevat. Thus I will follow the lead of Haspelmath (2000), who argues that optimality-theoretic analyses are often incomplete, because they fail to motivate their constraints. Haspelmath restricts his discussion to phonology and syntax, where it is still possible just to postulate a set of constraints, and get away with it, but people working in semantics and pragmatics 316 Buoyancy and Strength (especially pragmatics) tend to be less accommodating, as will be demonstrated in the following. Speaking loosely, presuppositions are interpretative elements that seek to have wide scope. This is but a loose way of speaking because presupposition projection is a pragmatic phenomenon, which has little to do with scope taking in the grammarians' sense of the word, but it serves well enough as a rough and ready characterization of what the projection problem is about. A more accurate, though less straightforward, description is the following: if a presupposition ip is triggered within the syntactic scope of any expression a, it will typically though not invariably seem as if ip is not affected by the presence of a, at all. To illustrate, if someone utters (ia), he will generally be understood as implying that (ib) is true, which is to say that the presupposition triggered by the factive verb (that the dean is a woman) appears to ignore the presence of the modal expression and the negative. (i) a. Perhaps Fred does not know that the dean is a woman, b. The dean is a woman. Why is it that presuppositions tend to take wide scope? The sad answer is that, at present, we do not know. I am not aware of any theory of presupposition projection that has a fully explanatory account of why presuppositions behave the way they do. Projection theories appear to fall into two classes: they either do not work or else they are forced to postulate, under some guise or other, that presuppositions tend to take wide scope. This may seem like a hopelessly embarrassing situation, but it is not: it is embarrassing, but not hopeless. For it can hardly be doubted that the workings of presupposition projection are better understood now than they ever were. There is a broad consensus about the mechanics of presupposition projection, and most people working in the field would agree, I believe, that the outlines of a solution to the projection problem have become reasonably clear. Yet the Big Why question still remains, and in the following pages I will first explain how it continues to haunt modern-day attempts at dealing with presupposition projection, and then consider at some depth an answer that has been suggested recently. Let us start out from one of the landmarks in the literature. Heim's 1983 paper is an attempt to show that 'presupposition projection is an epiphenomenon of the laws governing context change . . . ' (116). Heim presents a dynamic semantics which defines the meaning of an expression in terms of the effects it has on the context in which it is being used. Presuppositions, on Heim's account, are definedness conditions. If a ' Although Heim uses these words to characterize the motivation behind Gazdar's work, it is evident that, in the paper under discussion, she makes his aim her own.