It seems inconsistent to model boundedly rational action choice by assuming that the agent chooses the optimal decision procedure. This criticism is not avoided by assuming that he chooses the optimal procedure to choose a procedure to ... to choose an action. I show that, properly interpreted, this regress, continued transfinitely, generates a model representing the agent's perception of all his options including every way to refine his perceptions. In this model, the agent surely must choose the perceived best option. Hence it is not inconsistent to model limited rationality by assuming that the agent uses the "optimal" decision procedure.