A basic unsolved problem in science is that of understanding learning, the process by which people and machines use their experience in a situation to guide future action in similar situations. The ideas of Piaget, Pavlov, Hull, and other learning theorists, as well as previous heuristic programing models of human intelligence, stimulated this attempt to achieve artificial intelligence by building an "artificial infant" which would "grow up" in much the same way as a human child. A computer program called INSIM1 is described which models simple forms of learning analogous to the learning of a human infant during the first few weeks of his life, such as learning to suck his thumb and learning to perform elementary hand-eye coordination. The program operates by discovering cause-effect relationships and arranging them in a goal tree. For example, if A causes B, and the program wants B, it will set up A as a subgoal, and then work backward along the chain of causation until it arrives at a subgoal which can be reached directly. The core of the learning program is an "experience driven compiler" in which causality is detected by statistical correlation. This work is discussed as it relates to fundamental scientific issues, to other research on learning, and other research on artificial intelligence. Some proposals for future development are discussed at length. (JY)