The hypothesis that role-playing scenarios specifying full rhetorical contexts were a superior means of eliciting valid writing samples for the purpose of assessing compositional skills of prospective Florida teachers was tested. In addition, topics for use on the initial administrations of the Florida Teacher Competency Examination's writing subtest were developed and validated. Six essay topics selected by a panel of validators were cast in three different presentation modes according to degree of specification of rhetorical context, or "informational load (IL)." A low IL presented the topic in a brief phrase leaving the writer free to make decisions about audience, purpose, form, and tone without guidance. A moderate IL gave the writer an orientation to the task without including specifications. A high IL gave a full rhetorical context. The six topics in the three modes, 18 combinations in all, were administered randomly to a sample of college education majors. Evidence gathered in statistical and rhetorical analyses points clearly at the moderate IL as the presentation mode most likely to stimulate the best writing of large number of examinees. (RL)