Research on recasts has gotten increasing attention from SLA researchers ever since 1990s. However, the studies on the effectiveness of recasts had inconsistent findings and are still debatable in what way recasts facilitate L2 learning. This study aims to examine the effects of recasts on English question production by adopting a pretest, posttest, and delayed posttest design. Grade eleven students participated in the study and were grouped into three developmental stages of the target structure, the intended framework. The transcribed interactional tasks were coded in terms of recasts, noticing of recasts, perceptions of recasts, learner uptake, primed production, developmental stages of questions, and upgrade of developmental stages, respectively. Most students exhibited substantial preferences for recasts over other types of feedback in their interview data, with more than half of the students in each group reporting noticing the recasts in their learning journals. Primed production played a better role than uptake. Successful uptake was somewhat infrequent and was not significantly correlated with any group's noticing of recasts. It had a significant impact on the question development of lower-level and intermediate-level students but not on higher-level students' accuracy of question production. In contrast, it was revealed that primed production of question forms was strongly correlated with noticing recasts for all groups, and it was facilitative of subsequent L2 development in all groups. It concludes that recasts provided in dyadic interactions have facilitated the English question production by Ethiopian EFL learners at different stages.