Comparing Context Updates in Delineation and Scale Based Models of Vagueness

Although there exists a multitude of different approaches to vagueness in linguistics, there seems to be a general agreement that vagueness involves one or another kind of context dependence. E.g. Lewis [17] gives a high level overview of the use of contexts in natural language and illustrates that they are essential for modeling the shifting of vague standards. Therefore a vague sentence cannot be interpreted in isolation from the rest of a discourse; instead its truth conditions depend on—implicit or explicit— assumptions made by the conversationalists so far. Models of vagueness can be roughly divided into two groups: scale based and delineation models. Whereas scale based models explicitly model vague standards as cut off points on some appropriate scale, delineation models take another route. Vague adjectives are treated as boolean predicates whose extension crucially is context dependent. Scale based models are nowadays more popular in linguistics; there is much work going on in exploring connections between the scale structure of gradable adjectives and their use in natural language. Delineation models, on the other hand, have their roots in the philosophy of vagueness and typically make use of supervaluation [5]. The general aim of this contribution is to compare these two kinds of approaches in terms of their ability to model the notions of context and context update and to elucidate where their particular strengths and weaknesses lie in this respect. For this purpose, two concrete models are picked, namely Barker [3] as an example for a scale based approach and Kyburg and Morreau [15] as an example for a delineation based approach. Finally also Shapiro’s [19] account of vagueness in context is presented and we point out how to combine it with Kyburg and Morreau’s approach. Shapiro himself is not a linguist, but his model explicitly refers to conversationalists taking part in a conversation during which vague standards are shifted. The main result is to show in which situations Barker’s and Kyburg and Morreau’s approaches make the same predictions and which aspects cannot be modeled properly in either model. As an example Kyburg and Morreau’s model is clearly superior when the hearer is confronted with new information which requires the shifting of vague standards such that previously established information has to be retracted. Barker however does not consider such situations; once vague information is added to the current context it is never invalidated. On the other hand, Barker’s model provides more flexibility when it comes to predicate modifiers such as