Self-evidence is widely taken to be a status that marks propositions as capable of being justifiedly believed (and known) on the basis of understanding them. This paper explicates and defends that view. The paper shows that the broadly linguistic kind of understanding implied by basic semantic comprehension of a formulation of a self-evident proposition does not entail being justified in believing it; that the kind of understanding adequate to yield such justification is multi-dimensional; and that there are many variables partly constitutive of such understanding—all having philosophical interest in themselves—that a theory of self-evidence must account for. The paper also shows why self-evident propositions need not be obvious, need not be unprovable, and, far from being beyond dispute, can be a subject of rational disagreement. The concluding section shows how knowledge of self-evident propositions is possible even if, on the one hand, their elements are abstract and causally inert and, on the other, beliefs constituting knowledge must meet both causal and reliability conditions connected with their truthmakers.