A decision maker with incomplete preferences faces a choice problem over a finite set. Learning occurs when the decision maker is able to order additional pairs of elements from the choice set. When learning satisfies a set of four axioms, we establish that the set of learning processes (the different ways in which learning occurs) is equal to the set of order-extensions of the preference relation of an indecisive decision maker. Preferences settle on one (among finitely many) complete preference orderings in finite time. Heterogeneity in a cross-section of decision makers arises due to the different ways in which individuals learn, and the different stages of learning a person is at within a given learning process. Several applications of the framework are then discussed, including the measurement of indecisiveness, how learning occurs from social interactions, the utility representation of learning processes, and the implications of learning for axioms of revealed preference.