This study aimed to examine how intermediate-level learners of Japanese predict the content of the following sentences when they listen to sentences in Japanese, by using a test that is quite similar to SPOT. The results were as follows: (1) the size of the recognition load had no influence on the test score. (2) They probably predict the content of the following sentences incrementally based on a small unit, such as _bunsetsu_ (that approximately corresponds to a word plus bound morphemes), so that they can minimize the recognition load. (3) It was also confirmed, as has been done in the literature, that the intermediate-level learners of Japanese can make great use of predictions when they listen to the sentences.