Abstract

Russellian or singular propositions are very useful in semantics to specify "what has been said" by a literal and serious utterance of a sentence containing a proper name, an indexical or a demonstrative, or for modeling demonstrative thoughts. I3ased on an example given by S. Guttenplan, I construct a case showing that if our only option for modeling demonstrative thoughts is a singular proposition à la Russell, we run the risk of admitting infallible empirical (existential) beliefs. I defend the principle of the fallibility of our (first order) representations by appealing to Perry's notion of a relational mode of presentation that allows us to generalize the proposition which is the content of the perceptual belief in cases of hallucination or misidentification, so that there is no "immunity to error through misidentification" in the province of demonstrative thought.