In this paper, we distinguish two ways in which someone can be said to believe a proposition. In the light of this distinction, we question the widely held equivalence between considering a proposition true and believing that proposition. In some cases, someone can consider a proposition true and not properly believe it. This leads to a distinction between the conventional meaning of the sentence by which a subject expresses a belief and the content of this belief. We also question some principles of belief ascription, suggest a solution to a famous puzzle about belief and defend the unity of the semantic and causal aspects of beliefs.