Skip to Main Content
In order to study the structure and acquisition of perception and motor skills, we are simulating on a digital computer some features of a baby's sensorimotor development. The baby comes to be able to recognize and manipulate objects, taking into account their movements and other spatial relationships. He performs purposeful actions naturally described in terms of their effects on his environment rather than in terms of particular muscle movements. For instance, we say that he picks up his rattle rather than saying that he moves certain muscles because from our usual point of view, we care about the act as related to other acts. The movements could have been any of a large number of movements performed by different sets of muscles so long as they combined to produce the desired effect of picking up the rattle. We are trying to learn more about the sequence of development which brings about this purposive regulation of movements. There have been a number of different approaches to problems like these. One is to deal with elements at the neural network level adjusting connection strengths of thresholds, while another is to write computer programs in which symbols may designate complex behavioral acts. We are working somewhere between these levels.