Summary: Tarski's notion of logical consequence is viewed as a special case of the more general notion of being a theorem of an axiomatic theory. As was recognized by Tarski, the material adequacy of this definition depends on having the distinction between logical and nonlogical constants, but we find Tarski's analysis persuasive even if we do not agree on what constants are logical. This accords with the view put forward in this paper that Tarski indeed captures the more inclusive notion of theoremhood in an axiomatic theory. The approach to logical consequence via axiomatic theories leads us to grant centrality of inference schemes rather than to full-fledged arguments and to view the logically valid schemes as a subclass of generally valid schemes.