Possible-worlds analyses of belief attribution have faced insuperable problems dealing with necessities and impossibilities since their inception 6 decades ago. I present a new analysis of belief attribution that overcomes these problems. A sentence of the form a believes that p is argued to be true if and only if p can be inferred from beliefs of a combined with a limited range of beliefs of the attributor using a restricted syntactic inference. Crucial to success of the analysis is adoption of a novel I-semantics that takes interpretation to differ from person to person, formalized using multiple possible interpretation functions but only one possible world, the actual world. In addition to being empirically superior to all known alternatives, the proposed analysis also gives a uniform account of de dicto, de re, de qualitate and de translato interpretations based on only a single source of ambiguity, induced by raising into a higher clause.
A published version of the paper this talk is based on can be found here: https://doi.org/jgjj