ABSTRACT Three different analyses were conducted on the data from oral interviews with six child learners of Japanese, using a newly developed interview procedure called Pro‐I. The first analysis to assess the effectiveness of different tasks suggested that tasks sensitive to the students' understanding of certain critical syntactic structures and tasks manageable for children appropriate to their level of cognitive development were effective in discriminating students' language development levels. Also, free‐form tasks were effective to elicit ability of students with higher proficiency to its fullest, and tasks with motivating topics could generate drastically better, more expressive performance. This suggests that Japanese FLES assessments should include different types of tasks. In the second analysis, in which each student's language behavior was closely examined, it was observed that high‐level students tried to produce a greater quantity of Japanese with more complexity rather than more appropriate sentences, while most other students used one‐word answers extensively. Also, a high‐level student's performance exhibited monitoring and control over his own Japanese production. The third analysis found that young learners' phrase particle use was characterized by scarcity of misuse, omission, and pauses, suggesting children's tendency to learn multiwords as formulaic chunks. Repeatedly revealed characteristics unique to child foreign language learning throughout three analyses imply that we should not simply apply the instruction and/or assessment designed for adults to young learners and should accumulate knowledge necessary to develop instruction and assessment appropriate for children.