Cohen's [1] refinement rules provide a flexible mechanism for introducing intentional background knowledge in an ILP system. Whereas Cohen used a limited second order theorem prover to implement the rule interpreter, we extend the method to use a full Prolog