Four Levels of Learning

From P2P Foundation
Jump to navigation Jump to search

As proposed by Gregory Bateson.


By Jonas Andersen:

"Here is a brief overview of the logical categories of learning and communication, which he calls ‘Learning Zero-III’.

Learning Zero – Causality: The best description of categories of learning and communication starts with what it is not. In this case a strictly causal response to a certain stimuli is not learning or communication. When I let go of the coffee mug in my hand, it will inevitably fall on the floor leaving a nasty stain. I cannot condition the mug to hover or levitate to a soft landing on the table. This is not leaning. Or communication.

Learning I – Linearity: Gregory Bateson exemplifies Learning I by describing Pavlov’s famous psychological drawling dogs experiment. Every time he fed the subject dogs, he first rang a big brass bell. Every time he rang the bell, he recorded the dogs’ saliva excretion. At first the dogs drawled only at the exact time they were fed and god a scent of the food, but after a while the dogs connected the ringing bell with feeding time and began drawling when the bell rang. Eventually he had conditioned the dogs to start drawling whenever he rang the bell regardless of whether it was feeding time or not. What Pavlov did was conditioning a certain input-output equivalence in the drawling system of the dogs. They were taught that if the bell rings then dinner will be served shortly. This is something algorithms of digital media do very well. Correction of response in linear causality is consequently what Bateson calls ‘Learning I’.

Learning II - Complexity: Learning II has had many names including ‘Deutero Larning’, ‘set learning’, ‘learning to learn’, and ‘transfer of learning’. It is change in the process of Learning I, e.g. a corrective change in the set of alternatives from which choice is made (Bateson 1972 p. 293). So when a change in the set of possible alternative responses is learned it will fall under the category of Learning II. In other words if the dogs learn to learn to correct their response to a given stimulus they will have classified the contexts of learning I. Or more precisely, if Pavlov’s dogs were placed in a conditioning experiment under completely different circumstances and needed less time to correct their response to the stimulus, this would be Learning II. Learning II is learning to learn.

Learning III - Hyper-complexity: Learning III is a change in the process of Learning II, e.g. a corrective change in the system of sets of alternatives from which a choice is made. It works much as Learning II but on a higher level of abstraction. Bateson points out that this level of learning is never achieved by dogs and very rarely by humans, but as we will see a bit further on larger social systems have much greater capacity for handling the complexity at level III than any human." (


Gregory Bateson paper on ‘The Logical Categories of Learning and Communication’ first published in august 1968 at the Conference on World Views held by the National Institute of Mental Health (Bateson 1972 p. 279-308).