This paper offers an interim analysis of written feedback on a first draft of a PhD thesis. It first looks at two sources of data: in-text feedback and overall feedback. Looking at how language is used in its situational context, we then coded the feedback and developed a model for analysis based on three fundamental functions of speech: referential, directive and expressive. It was found that expressive feedback benefited the supervisee the most. The interaction between the supervisor and the supervisee played an important role for the induction of the supervisee into the academic community, and suggests a peer-to-peer model in PhD education. Finally, this paper suggests the possibility of developing a taxonomy of good feedback practices in postgraduate supervision practice in Higher Education. (Contains 2 tables and 1 note.)