By performing a series of five experiments with two subjects, several aspects of one of the subject's behavior in solving chess problems were found to be predictable, and a model was developed to explain this predictability. The heuristics used in this model may be applicable in developing future computer programs for chess play.