Landau (2016b) draws on data from person mismatches in partial control to argue that agreement is (at least sometimes) a PF (Phonological Form) phenomenon. The argument depends in part on the premise that there is a reading of sentences such as They wanted to prepare themselves and then meet for debate that corresponds to a structure in which the VP meet for debate has a controlled PRO subject that is semantically first‐person but morphologically third‐person. I argue that according to Landau's own assumptions, PRO in such sentences is not semantically first‐person at any level of representation: Landau's premise relies on a problematic conflation between PRO and the output obtained by applying the group operator to PRO in Landau's approach to partial control. Consequently, Landau's argument for the PF status of agreement does not go through.