Theory posits that skilled human movements, from the simplest pointing movements to more elaborate multi-gesture movements, are directed by "motor programs." Although we are seldom aware of these programs, they act to structure commands to muscles so that movements can be made with minimal reliance on slow, feedback- based, error-correcting processes. The idea of motor-program generalization stems from the computer-program analogy and suggests that motor programs have formal parameters which take on situation-specific values when a movement is made. The focus of this research is whether and how motor programs can be generalized so that a single representation is sufficient to control related, although perhaps novel, variations of a movement. The research will address these issues in three empirical domains, handwriting, simple drawing movements, and aimed movements made with excess degrees of freedom. The work on handwriting will examine changes in a highly overlearned skill when subjects are constrained to deviate from normal performance conditions in one or more ways: e.g., size, speed, or effector. This research will identify the type and magnitude of constraints that necessitate a shift from the motor- program-based, control strategy to a feedback-based strategy of movement approximation. Using drawing of simple figures, the second line of research will assess the nature and degree of transfer that occurs when people occasionally are asked to produce, under a particular set of constraints, a figure which they are learning to produce, by practicing with feedback, under a different set of constraints. The patterns of learning transfer in these experiments will provide evidence about motor program generalization and will provide a way to explore the possibility that motor programs have a hierarchical structure. The third line of research will investigate the strategies and constraints used to resolve the planning ambiguity that occurs when excess degrees of freedom are available to perform aimed movements. Unlike the first two lines of research, which assume an analogy to motor-program generalization based on the modification of pre-existing information, in this research motor programs and their generalization will be modeled as the result of a process of dynamic constraint optimization. The insights about the structure and function of motor programs gained from this research have potential for application to the training of skills requiring coordinated movement, the diagnosis and treatment of movement disorders, and the planning and representation of movements in robotic systems.