Mathematics teacher educators' (MTEs') written feedback is an instantiation of teaching practice and provides information to support the developing understanding of prospective teachers (PTs), yet no research has been undertaken to characterize such feedback. Research on feedback forms, including written feedback, has focused on information provided to learners that draws attention to different components of their performances or understandings, such as elements of the task and processes used to complete it. While no models of feedback or feedback systems prescribe what feedback should contain, factors that researchers have identified as impacting performance or understanding under some conditions have been identified. To explore MTEs' written feedback as an instantiation of practice and to characterize such feedback, we used a self-study methodology to analyze three MTEs' written feedback to PTs regarding their responses to students, including the process of attending to and using learners' mathematics in responses to written work. Findings showed MTEs used some practices described as effective feedback by building on to PTs' feedback to mathematics-learners. Yet they also provided feedback described as less than effective, such as redirecting PTs to attend to elements in students' work not addressed by the PTs but perceived by MTEs as important to address. In addition, findings illustrated that MTEs provided little feedback on PTs' self-regulation. We argue that if MTEs' written feedback is meant to serve as an instantiation of practice as well as provide information to PTs about their feedback practice, MTEs should explore their feedback so that the model provided operates as they intend.