With the development of the information society, the number of labors involved in manual or physical work has decreased and the number of labors involved in intellectual work has increased. As a result, management of the state of intellectual concentration has been attracting attention to improve work performance. To do it, it is necessary to clarify what causes changes in the state of intellectual concentration. Therefore, this study focuses on ACT-R, a cognitive architecture that can simulate human cognition. We focus on the fact that the time required for cognitive behavior varies even in the state of intellectual concentration and aim to construct a fundamental model that represents the mechanism of change in the state of intellectual concentration by reproducing it using ACT-R. We conducted an experiment where eye movements and response times in an intellectual concentration state were measured using a simple addition task. Furthermore, by introducing the concept of the short-term pause state and the long-short-term pause state, we constructed a model that reproduces the cumulative frequency distribution of the variation in response time among the measurement results with high accuracy using ACT-R.