International Encyclopedia of Systems and Cybernetics

1. "The process of acquiring knowledge and skills through practice, study or information" (UNESCO-UNEP Glossary, 1983, p.17).

2. "The general class of processes by which the behavior of an animal comes to depend not only of the environmental changes immediately preceding it in time, but also on events which have occured in the related parts of the environment in the more remote past" (J.W. PRINGLE, 1956, p.93).

3. "The retention of adaptive response patterns for subsequent utilization, thus abbreviating the trial and error process (D.T. CAMPBELL, 1975, p.17).

4. "The reduction in the randomness of behavioral elements" (R.W. FULLER & P. PUTNAM, 1967, p.104).

These very different definitions (and many others have been given) show the complex nature of learning.

Learning is a very general obility of living systems. Basically it "… consists in the formation of new neural patterns, i.e. in establishing permanent connections among neurons or facilitating ephemeral (but repeatable) interconnections among them. Some connections may be formed accidentally (at random); if valuable, they have a chance of becoming established or recurrent" (M. BUNGE, 1979, p.133).

This has been demonstrated in lower organisms such as cockroaches or slugs. Low level learning, at least, does not even depend on the existence of the neo-cortex, nor forebrain, as mentioned by S. WALKER (1983).

PRINGLE makes an important comment to his definition: "The modification of present behavior by past events is called learning when those events have occured during the lifetime of the individual, and instinct. when the events have occured outside this period of time. In the former case the past events are supposed to have left some trace in the animal (usually in its nervous system), whereas in the latter case the organization of the animal responsible for the observed response is supposed to be innate: that is to be contained fully in the material substance of its inheritance. This distinction is clear-cut theoretically, although great difficulty has often been experienced in assigning to one or other case the particular aspect of behavior which is the subject for study" (Ibid., p.74).

W.R. ASHBY stresses that "learning usually changes behavior from a less to a more beneficial, i.e. survival promoting, form" (1960, p.3- 4).

He adds that "The nervous system is well provided with means for action. Glucose, oxygen and other metabolites are brought to it by the blood so that free energy is available abundantly… The nervous system, then, possesses almost unlimited potentialities for action".

But this does not explains why learning is beneficial: "We are concerned primarily with the question why, during learning, behavior changes for the better; and this question is not answered by the (mere) fact that a given behavior can change to one of lesser or greater activity. If a dynamic system is allowed to proceed to vigorous action without special precautions, the activity will usually lead to the destruction of the system itself" (p.4-5).

ASHBY's book "Design for a brain" is mainly trying to explain this basic question by developing a model of cybernetic acquisition of cerebral organization, with the appropriate constraints and heterogeneity.

As to FULLER and PUTNAM, they understand the learning process in a negative way "… rather than as the incorporation of order in a blank photographic plate" (p.104) and, through a random search process as "… the discovery of more and more relative dominances through interacting with (the) environment" (p.105).

R.L. ACKOFF stresses another important point: "Since learning can take place only when a system has a choice among alternative courses of action, only systems that are goal-seeking or higher can learn. If a system is repeatedly subjected to the same environmental or internal change and increases its ability to maintain its efficiency under this type of change, then it learns how to adapt. Thus adaptation itself can be learned" (1972).