Acquisition and Improvement of Human Motor Skills: Learning through Observation and Practice Wayne Iba Doctoral Dissertation Abstract 1991 Skilled movement is an integral part of the human existence. This is exemplified in a range of behaviors from concert violin performance, to picking up and drinking a glass of milk. A better understanding of motor skills and their development is a prerequisite to the construction of truly flexible intelligent agents. Existing computational models have mostly focused on low-level issues of controlling manipulators rather than on capturing skilled movements as conceptual units. The psychological literature provides very high-level abstract theories or low-level analysis of specific movement phenomena. Furthermore, the acquisition of skills is largely ignored in both bodies of work. In response to these issues, we present MEANDER, a computational model of human motor behavior, that uniformly addresses both the acquisition of skills through observation and the improvement of skills through practice. Meander consists of a sensory-effector interface, a memory of movements, and a set of performance and learning mechanisms that let it recognize and generate motor skills. The system initially acquires such skills by observing movements performed by another agent and constructing a concept hierarchy. Observed movements are parsed and stored internally as motor schemas. Two subsystems of MEANDER interact to allow observed movements to be recognized and stored skills to be executed. The OXBOW module is responsible for constructing and modifying the skill hierarchy according to the observed experiences. Given a stored motor skill in memory, the MAGGIE component will take the motor schema and cause some effector to behave appropriately. Errors in execution can be corrected through a closed-loop feedback control mechanism. All learning involves changing the hierarchical memory of skill concepts to more closely correspond to either observed experience or to desired behaviors. One can evaluate the effectiveness of a model in a number of ways. We evaluate MEANDER empirically with respect to how well it acquires and improves both artificial movement types and handwritten script letters from the alphabet. We also evaluate MEANDER as a psychological model by comparing its behavior to robust phenomena in humans and by considering the richness of the predictions it makes.