Stochastic models explaining the hierarchical organization and the learning potentialities of a neural net are presented. The elements of the neural net appear as states of chains with infinite memory. The ergodic properties of the net allow to explain naturally the hierarchical organization of neural networks with an abundance of branchings and stable states. The stochastic model is able to describe some properties of nets e.g. the memory, the adaptability, the storage of information, the sensitivity to initial conditions. Three frameworks of the time are outlined; rest, uniform time and continuously but purely singular time. The possibilities of using stochastic and non‐archimedean methods in higher nervous activity modelling are illustrated.