The very original concept of cognitive radio raised by Mitola targets at all the environment parameters, including those in the physical, MAC, and application layers, as well as the information extracted from reasoning. Hence, the first CR is also referred to as “full cognitive radio.” However, due to its difficult implementation, the FCC and Simon Haykin separately proposed a very simplified definition, in which CR mainly detects a single parameter, spectrum occupancy, and is also called “spectrum sensing cognitive radio.” With the rapid development of wireless communications, the infrastructure of a wireless system becomes much more complicated, while the functionality at every node is desired to be as intelligent as possible, say selforganized capability approaching 5G cellular networks. It is then interesting to reconsider Mitola’s definition and think whether one could, besides obtaining the on/off status of the licensed user only, achieve more parameters in a cognitive way. In this article, we propose a new cognitive architecture targeting multiple parameters for future cellular networks, which is one step further toward “full cognition” compared to most existing CR research. The new architecture is elaborated in detailed stages, and three representative examples are provided based on recent research progress to illustrate the feasibility as well as the validity of the proposed architecture.