"The learning nervous system is a deterministic system with a relativistic self-regulating organization that defines its domain of interactions in terms of the states of neuronal activity that it maintains constant, both internally and at its sensory surfaces, and that specifies these states at any moment through its functioning, and through the learning (historical transformation) itself. Consequently, it must be able to undergo a continuous transformation, both in the states it maintains constant, and in the way it attains them, so that every interaction in which new classes of concomitances occur effectively modifies it (learning curves) in one direction or the other. Since this transformation must occur as a continuous process of becoming without the previous specification of an end state, the final specification and optimization of a new behavior can only arise through the cumulative effect of many equally directed interactions, each of which selects, from the domain of structural changes possible to the nervous system in its structural dynamism, that which at that moment is congruent with its continued operation subservient to the basic circularity of the organism. Otherwise the organism disintegrates."
Humberto Maturana and Francisco Varela, Autopoiesis and Cognition.