Self-Explanation is considered to be one of the effective ways to elicit active knowledge construction in general domain. Although the effects of promoting self-explanation have been demonstrated in a variety of domains, there is some discrepancy on the effect on scientific conceptual learning between prior researches. To address this issue, the approach taken here was to develop new prompts based on SBF theory, in which complex systems were described in terms of function and behavior of the components, and to compare the new one (SBF prompts) with the one used in prior researches (Generic prompts). 47 students participated in an experiment and were randomly assigned to SBF prompts group, Generic prompts group, and control group (think-aloud without prompts). Results showed that the performance of SBF prompts group was better than that of control group on inference questions which refered to the function or behavior of components in the text. On the other hand, Generic prompts were not so effective for eliciting self-explanation inferences and for increasing performance on post-tests. In addition, protocol analysis reveled that a learning gain by SBF prompts was mediated by SBF-based explanations during learning. These results suggest that we have to use the prompts which require inferences and monitoring based on a SBF theory to promote scientific conceptual understanding.