On a revise-and-resubmit assignment in a large introductory History course, students were provided with feedback that was phrased either as questions, statements, or imperatives. This study examines which form was most likely to lead to improvement in the students' writing. Generalized estimating equations (GEE) were used to analyze a data set comprising 669 individual pieces of feedback on 67 sets of papers. Researchers found that, overall, students were most likely to implement feedback phrased as imperatives and least likely to implement feedback phrased as questions, and that the likelihood shifted somewhat depending on which aspect of writing was being commented upon; the extent of change required; the students' past performance in the course; and the person providing the feedback.