A concept that all hierarchical levels of the nervous system are built according to the same functional principles is proposed. Each level is responsible for a discrete type or set of automatisms, is a learning system, and contains two distinct functional subdivisions: (1) a controller, a subsystem providing a governing set of rules or commands-a control law-that directs the action of the recipient of these rules-the controlled object; and (2) a model, a subsystem that generates a model of object behavior, i.e. afferent information flow expected from the controlled object. A control system such as this receives two types of afferent signals-initiating and informational. The difference between these signals is that a control system minimizes initiating signals during the realization of an automatism, i.e. a control neural network utilizes informational signals to compute the proper output that minimizes the initiating input signal. A mismatch or error signal, a type of initiating signal, is responsible for learning. Both the control law and the model can be adjusted during learning. The learning process starts when the error signal increases and stops when it is minimized. A network hierarchy is structurally and functionally organized in such a way that a lower control system in the nervous system becomes the controlled object for a higher one. This hierarchy leads to a generalization of encoded functional parameters and, consequently, the working space for each higher level control system becomes more abstracted. This is the reason why each hierarchical level within the control nervous system uses detectors specific for feature of the controlled object and the environment that match the control needs in order to obtain information about the current state of the object in the environment. Movement of information toward higher hierarchical levels also is accompanied by an increase in the duration of initiating signals within each control system. The ability to store a long prehistory of preceding events is considered as the mechanism that necessitated the invention of more complex and more rapid forms of learning such as operant learning, and made possible more complex multistep computational algorithms that require memorization of the results of previous intermediate computations. The functions of the cerebellum, the limbic system and the cortico-basal ganglia-thalamocortical loops are analyzed to illustrate the utility and applicability of this theoretical concept. Basal ganglia-thalamocortical loops are described as modeling, predictive loops, and their dopaminergic innervation as an error distribution system.