While existing studies have underscored the educational benefits of generating explanatory hypotheses (EHs) in response to unexpected outcomes, empirical research on the underlying mechanisms driving their effectiveness in science learning remains limited. Thus, this study aimed to empirically examine the effectiveness of generating an EH for incongruent results in enhancing science learning. Two studies were conducted with 208 students who faced incongruent results out of 578 fifth and sixth-grade students. Study 1 revealed through hierarchical regression analysis that generating an EH predicted enhanced science learning and uncovered the highest predictive power compared to other predictors, including prior knowledge and motivating variables. Study 2 explored the contribution of generating an EH for incongruent results to comprehending scientific concepts through think-aloud protocols and retrospective interviews. The results demonstrated that, unlike students who could not generate EHs and passively acquired information from the text, students who formulated EHs more actively organized information by verifying their hypotheses and paraphrasing the scientific concepts in the text during reading the learning material after generating their hypotheses. The study findings indicate science education direction for future research and broaden the discourse surrounding effective educational practices.