In recent Years, much of the efforts in developing Intelligent Tutors focused on student modelling. Student models aim at representing both correct and incorrect knowledge of the student in the particular domain of learning. This representation serves as the basis upon which the tutor can dispense remedial feedback. This paper presents a framework in which learner's misconceptions are diagnosed, and linked to some cognitive processes, such as ignorance, generalization, which are known to be involved in learning. This framework also enables the system to compute the correct feedback which has to be given to the learner in each case of misconception.