This multiyear study helps elucidate how the instructional practice of student-generated questions support learning in a blended classroom in science, technology, engineering, and mathematics subjects. Students designed multiple-choice pre-exam questions aimed at higher levels of learning, according to Bloom's taxonomy. Student-generated questions were edited by the instructor and then discussed by the students in the classroom and in an online forum. We tested the hypothesis that this intervention improves student learning, measured as student achievement on the exam following the intervention, and compared to student achievement on the traditional exam (prior to which a review session focused on instructor-led recitation of the key concepts). Following the intervention in all years, average grade on the post-intervention exam increased by 7.44%. It is important to point out that not all students benefited equally from this activity. Students who were in the 4th quintile (60% to 80%) based on the results of the 1st exam demonstrated the highest achievement improving their performance on average by 12.37% percentage points (measured as a score on the 2nd exam). Gains were not observed in the semesters when the intervention was not implemented. In this study we provided students detailed instructions on how to design questions that focus on testing higher levels of learning.