Stalnaker ’ s Thesis in Context

A popular form of contextualism concerning indicative conditional statements embraces the following three theses: (i) Indicative conditionals semantically express (i.e. can be used to assert the truth of) propositions. (ii) Which proposition is asserted by an utterance of an indicative conditional sentence sometimes depends on the context in which it is uttered. Moreover, (iii) which proposition is asserted depends, absent other sources of context sensitivity, on some piece of evidence or knowledge that is salient in the context of utterance (perhaps the utterer's evidence or some pooled piece of evidence being assumed by the participants of the conversation.) Contextualism holds a distinguished place in recent philosophy (see [18], [27], [14], [15]), and for good reason. It promises to answer a number of rather puzzling issues in the philosophy of conditionals – the apparent validity of 'or-to-if' arguments (see [18]) and the so-called 'Gibbardian stand-offs' [27] to name but a couple of examples. While it also has its dissenters 1 it is interesting to note that, by contrast, the context sensitivity of conditionals is all but taken for granted by linguists working within the framework of Kratzer's [10], which draws on an important connections between modals and conditionals (and in the former case at least, context sensitivity appears to be pervasive.) My aim here, however, is not to defend contextualism or its applications, but to show that contextualism can be put to work to shed light on another difficult issue, namely that of providing a theory that predicts our intuitive judgments about the probability of conditional statements. Starting with Lewis's original paper on this topic it has often been noted that contextualism provides a potential way around the so called 'triviality results'. However, despite some limited initial optimism regarding this project (see in particular Harper [8] and van Fraassen [26]), few philosophers still see this as a viable option. Most likely this is due to the fact that existing constructions either fall apart when one considers conditionals embedded within other conditionals or are not compatible 1 There is a long tradition rejecting thesis (i) of the contextualist program: see Adams [1], Edgington [5] and Bennett [3] for representative examples. More recently some theorists have attempted to accomodate the data keeping (i) but rejecting (ii) by adopting a form of relativism about the propositions expressed by conditional sentences; see Weatherson [28] for discussion of this approach.