In this paper, learning is conceptualized as a process of routinization of learners' actions. We begin by operationalizing the keyword "routine." Two categorizations of routines are then offered. First, we distinguish between "practical" and "discursive" routines, discuss the process of their historical co-development, and claim that over time, new discourses germinated in routines straddling different types of practical and discursive activities. Second, the distinction between process-oriented and product-oriented routines is introduced. The former type of routine is called "ritual" and the latter "deed" or "exploration," depending on whether the routine is practical or discursive. The claim is made that in the process of learning mathematics, the germinal routines, from which a discourse new to the learner is to emerge, are initially implemented as rituals. In the longer run, these routines are expected to undergo gradual de-ritualization until they become fully fledged explorations. The question of how to identify the gradual advances in this latter process is then addressed, and this is followed by a brief discussion of the ways in which teachers may support or hinder transforming rituals into explorations. We conclude by taking a critical look at relative advantages of conceptualizing learning as the process of routinization.