The purpose of this dissertation was to design, develop and field test an Intelligent Tutoring System (ITS) which I have called Process Oriented Subtraction-Interface for Tutoring or POSIT. POSIT is an Intelligent Tutoring System, developed on a microcomputer, and based on Anderson's (1982, 1987) ACT* model of learning. Unlike the tutoring systems that were developed by Anderson and his colleagues (Anderson, Boyle & Reiser, 1985; Anderson & Reiser, 1985) which focused on the tutoring of students in the context of problem solving, this system focuses on the tutoring of a cognitive skill-subtraction of whole-numbers. Because ACT* theory explicitly describes the interaction of declarative and procedural knowledge (procedural knowledge is dependent on declarative knowledge), this learning theory is ideally suited to the learning of a specific procedure. Further, other "intelligent" systems which have been applied to subtraction (Brown & Burton, 1978; Ohlsson & Langley, 1985; Young & O’Shea, 1981) tend to focus on the answers to subtraction exercises (product oriented). POSIT, on the other hand, is an interactive system that determines errors made by the child as the child attempts to solve subtraction problems. Another difference with previous systems is that POSIT has a teaching component. Other systems assume that instruction has been given at some other point in time prior to the use of the system. One final difference between POSIT and other systems is that it was developed with consideration of the diverse student population that is found in schools. Design decisions were based on the cognitive needs of low-, average- and high-achieving students. Such notions as reading level and complexity of the task were considered with regard to low achieving students. POSIT's ability to allow for a wide variety of algorithms was a consideration with regard to high achieving students and would also benefit students from all levels. The results of the field test of POSIT indicate that the error model used by POSIT was very successful (76% accurate, with potential to improve to between 80 to 90%). In addition, students appear to learn from the system as measured both on the system as well as on a paper and pencil transfer test. / Ed. D.