The nature of the learning process of decision-making systems is conceptualized, and the process is formulated and explored in the framework of mathematical general systems theory. Learning is understood as a process by which the decision-making system acquires knowledge about its decision situations to increase the potential of its decision-making ability. First, learning is formulated as a generating process of a sequence of higher-order decision situations, then, after introducing concepts of solutions for the decision situations it is shown that as the order of the decision situations increases their solution sets expand. Furthermore, it is proved that there are certain limitations on the expansions and these are identified. During the generating process of higher-order decision situations, the order of the decision situations increases and they structurally change. In order to explicitly describe the structural changes of the decision situations, a category-theoretic framework is adopted. Then, the limitations of the expansions of the solution sets are expressed as categorical limits.