This paper describes an attempt to follow up the performance of foreign graduate students in the College of Engineering, University of Michigan, and discusses whether the scores they obtained on the Michigan Battery, with special reference to the composition test, predicted success or lack of success in their first term subjects. No significant correlation was found between students' cumulative grade point average and their Michigan Battery scores. A comparison was undertaken of the underlying requirements involved in writing for the composition test and writing for the majority of engineering assignments. It was found that the requirements differed considerably and that the main difference was in the amount of material the student was required to create in the Michigan composition test. What engineering readers look for is how well the student understands concepts, and relationships between concepts; whether steps are arranged in the right order and whether the work is accurate; whether the student arrived at the right answer; and whether the report was presented in a professional way. It was concluded that the proficiency test composition task should assess the student's ability to present information by providing the student with some sense of audience and purpose, and raw material to organize. The criteria for judging such writing should be open for discussion. (AMH)