This paper builds upon the analysis of Orphanides and Wilcox (1996) to evaluate optimal anti-inflation policy under a broader set of circumstances than considered in their work. We consider a monetary authority with two instruments--the funds rate and the discount rate--with the distinction that only movements of the latter are 'credible' alterations of the Fed's policy stance, reflecting reputational effects. The public forms expectations of inflation given realized inflation and the expected progress toward lower inflation, as evidenced by credible policy moves. Optimal policy is formulated in a stochastic, dynamic setting of the Tinbergen-Theil framework. The presence of a "cost-of-change" penalty on the sequence of discount rate adjustments generates expected trajectories for targets and policy instruments which differ considerably from those lacking such a penalty.