This study investigated how experimental manipulation of working memory load affected children’s input information retainment. 76 children (4;5-6;4, M=65 months, SD=6.61) were participated in our experiment. We conducted sentence repetition task for cleft sentences to examine children’s input retainment and set two conditions in which mora length of nouns and verbs in stimulus sentences were controlled: high working memory load condition (high WM load: 24 morae) and low working memory load condition (low WM load: 16 morae). As a result, children who repeated stimulus sentences perfectly in low WM load condition could not repeat case-markers correctly in high WM load condition. On the other hand, children who repeated stimulus sentences perfectly in high WM load condition repeated almost all cleft sentences in low WM load condition. Thus, working memory load affected children’s performance of sentence repetition. These results suggested that case-marking errors, which had been demonstrated in various sentence comprehension tasks, would be originated in input information retainment.