Research on intelligent tutoring systems has mainly concentrated on how to reduce a cognitive load which a student will bear in learning a domain. This load reduction approach contributes to facilitating his/her learning. However the approach often fails to reinforce the student's comprehension and retention. Another approach to tutoring is to apply a load to him/her purposefully. In this paper, we present a framework for cognitive load application and describe a demonstration system. The framework imposes a load on a student who tries to understand an explanation. The important point toward the load application is to provide the student with an optimal load that does not go beyond his/her capacity for understanding. This requires controlling the student's load by means of explanations. In order to implement such load control, it is necessary to estimate how much load the explanation imposes on his/her understanding process. The load estimate depends on his/her understanding capability since the same explanation imposes a different load according to the capability. Therefore a student model representing his/her capability is required. This paper shows how our system accomplishes a proper load application by generating explanations with the load estimate.