Abstract

In recent years Russell´s view that there are singular propositions, namely propositions that contain the individuals they are about, has gained followers. As a response to a number of puzzles about attitude ascriptions several Russellians (as I will call those who accept the view that proper names and indexicals only contribute their referents to the propositions expressed by the sentences in which they occur), including David Kaplan and Nathan Salmon, have drawn a distinction between what proposition is believed and how it is believed.1 While it is generally agreed upon among Russellians that this distinction needs to be drawn there is considerable disagreement as to what exactly the distinction amounts to and what role the what and the how should play. The most plausible option seems to be to not assign a semantic value to the cognitive role played by the name or indexical in the sentence assented to. But recently Mark Richard2 has attempted to make the cognitive role affect truth values by building it into the truth conditions of belief reports.3 I will argue that Richard’s attempt fails to satisfy our pretheoretical intuitions about the sharing of beliefs. Furthermore, and more surprisingly, I will argue that Richard’s theory makes it virtually impossible for us to judge whether or not most belief reports are true or false, since doing so would involve what I call RAM probing, which would require viewing elements that are essentially private.