Suppose a child were playing in a certain way, and a stranger appeared and began to scold and criticize. The child would become frightened and disturbed and try to escape. But if, in the same situation, the child's parent arrived and proceeded to scold and criticize, the result would be different. Instead of being frightened, the child would feel guilty and ashamed, and instead of trying to escape, the child would try to change what it was doing, in attempts to seek reassurance and approval.
I suspect that these two scenarios engage different learning mechanisms. In the encounter with the forbidding visitor, the child might learn I should not try to achieve my present goal in this kind of situation. But when scolded by someone to whom the child is attached, the child might learn I ought not to want to achieve that goal at all! In the first case, it is a matter of learning which goal to pursue in which circumstance; in the second instance, it is more a question of what goals one should have. If my theory is right, the presence of the attachment-person actually switches the effect of learning over to different sets of agents. To see the difference, let's make a small reformulation of the concept of a difference-engine to represent three different kinds of learning that an infant might use.
In the case of ordinary forms of failure or success signals, the learner modifies the methods used to reach the goal. In the case of fear-provoking disturbances, the learner may modify the description of the situation itself.
In the case of attachment-related failure or reward signals, the learner modifies which goals are considered worthy of pursuit.
So far as I know, this is a new theory about attachment. It asserts that there are particular types of learning that can proceed only in the presence of the particular individuals to whom one has become attached.