Several measurement theorists have convincingly argued that methods of writing test questions, particularly for criterion-referenced tests, should be based on operationally defined rules. This study was designed to examine and further refine a method for objectively generating multiple-choice questions for prose instructional materials. Important sentences were selected from a prose passage in a science text and these sentences were transformed into questions. Several variations of sentence transformation rules were used to create tests given to 273 college and high school students before and after they read the passage. Item difficulties (percent correct) for each type of item formed the basic data of the study. The study concluded that the method of selecting the "question word" (a noun or adjective) in the sentence has a crucial role in determining the pattern or pretest and posttest item difficulties of the resulting question. Also, the methods of item writing used in the study were found to be feasible and to be free from the item-writer differences that typically are found in traditional item-writing methods. (Author)