Quantitative data obtained during periodic clinical visits are merely “snapshots” of an individual's physiologic status or performance. The limitations of such periodic methods with regard to detecting transient events and in providing a true indication of global behavior (e.g., in response to a rehabilitation program) have given rise to long-term, portable monitoring instruments. This paper describes a new approach and device designed to objectively and unobtrusively monitor the behavior associated with human physical activity. The approach is based on the recognition that complex dynamic situations are frequently characterized by a single word (“walking”) suggesting that: (1) complexity can be reduced by looking for specificactivity states, and (2) each unique state is communicated by some combination of features which can be derived from measurable dynamic or static characteristics. The prototype Activity State Detection and Monitoring System (ASDMS) developed with this concept is capable of detecting and logging sitting/lying (considered as one state), standing, walking, and running activity states with a 5-sec time resolution over long periods. Evaluation studies demonstrate performance comparable to that of a human observer.