The way in which a theorem prover handles Definitions can have a significant effect on its performance. Many first-order clause form theorem provers perform badly on theorems such as those from set theory that are proven largely by expanding Definitions. The technique of using replacement rules permits automatic proofs of such theorems to be found quickly in many cases. We present a refinement of the replacement rule method which increases its effectiveness. This refinement consists in recognizing which clauses are obtained from first-order Definitions.