What is learning, anyway? That word is certainly hard to define. The child in our Block-Arch scenario has found one way to learn one sense of what some adults mean by arch. But we can't assume that the same kinds of processes are involved when we learn to recite a poem, to use a spoon, and to tie a shoe. What happens when a person learns to read, learns to add numbers, learns a new language, learns to anticipate the dispositions of a friend, or learns to build a tower that will stand? If we tried to find a single definition for learning to span so many kinds of processes, we'd end up with some phrase too broad to have much use — like this:
Learning is making useful changes in the workings of our minds.
The problem is that we use the single word learning to cover too diverse a society of ideas. Such a word can be useful in the title of a book, or in the name of an institution. But when it comes to studying the subject itself, we need more distinctive terms for important, different ways to learn. Even that one Block-Arch scene reveals at least four different ways to learn. We'll give them these new names:
Uniframing combining several descriptions into one, for example, by observing that all the arches have certain common parts.
Accumulating collecting incompatible descriptions, for example, by forming the phrase block or wedge.
Reformulating modifying a description's character, for example, by describing the separate blocks rather than the overall shape.
Trans-framing bridging between structures and functions or actions, for example, by relating the concept of arch to the act of changing hands.
These words will be explained in the sections that follow. It seems to me that the older words used in psychology — such as generalizing, practicing, conditioning, memorizing, or associating — are either too vague to be useful or have become connected to theories that simply aren't sound. In the meantime, the revolutions of computer science and Artificial Intelligence have led to new ideas about how various kinds of learning might work, and these new ideas deserve new names.
Our Block-Arch scenario is based on a computer program developed by Patrick Winston in 1970. Winston's program required an external teacher to provide the examples and to say which of them were arches and which were not. In my unprogrammed version of this, the teacher has been replaced by the concern of some agency inside the child to account for the emergence of that mysterious Hand-Change phenomenon: why do certain structures force you to let go of the toy car, while other structures don't? We thus assume that the child is led to learn for itself in order to account for strange events. One might complain that it only makes learning harder to explain, to make it depend upon the child's curiosity. But if we are ever really to understand how our minds grow, we must first face reality: people just don't learn so well unless they're interested or concerned. The older theories of learning and remembering never got very far because in trying to oversimplify, they lost essential aspects of the context. It wouldn't be much use to have a theory in which knowledge is somehow stored away — without a corresponding theory of how later to put that knowledge back to work.