The idea that impossibilities have a role to play in a semantic analysis of validity is becoming widely accepted on various grounds, including grounds of relevance. I argue that this is a mistake, that it has led to various foundational objections to relevance logic, and that these objections are avoidable given a semantics that clearly distinguishes two types of conditional or inferential fallacies, namely, those concerning truth preservation from those concerning a relation between content or subject matter. I argue that we should avoid the use of impossibilities employed in the standard Kripke semantics for relevance logic in favor of a coarse-grained notion of content, and argue for the benefits of such an approach. Since this discussion generalizes to more than relevance logic, I discuss the the role that impossibilities might have to play more generally in logical analysis.