We developed a system for detecting sentences considered difficult for Japanese language learners. This system will be a part of a text difficulty evaluation system, and is able to detect three sentence patterns; te-form and masu-form used as the gerundive form, and noun modification. We applied some rules to the results obtained by using a morphological analysis system JUMAN and a sentence structure analysis system KNP. It was found that this system was able to identify these three sentence patterns accurately. In cases where it is difficult to determine whether the text being analyzed includes any of the above sentence patterns, the system returns an unassertive judgment and leaves the decision to users.