In conventional written corrective feedback (WCF) practice, teachers spend an inordinate amount of time identifying every error in student writing. Research evidence suggests that such a comprehensive WCF approach is both undesirable and ineffective. Recent research has shown that focused WCF, where teachers respond to errors selectively, is a good practice since it is more manageable and less discouraging for students. Much of existing WCF research on focused WCF, however, has adopted the experimental or quasi-experimental design, involving a very small number of error categories, without paying attention to the real-life classroom conditions in which WCF takes place. To fill the research gap, the present study investigates how two secondary teachers select target language features for focused WCF and the issues that arise from their WCF practice. The paper provides pedagogical implications that contribute to our understanding of how teachers can go about selecting errors for focused WCF in authentic second language writing classrooms, as contrasted with experimental classrooms.