The technology of transforming sentences from prose instruction into test questions was examined by comparing two methods of selecting sentences (keyword vs. rare singleton), two types of question words (nouns vs. adjectives), and two foil construction methods (writer's choice vs. algorithmic). Four item writers created items using each combination of the methods listed above, for a 1,000 word passage. A sample of 273 college and high school subjects were randomly assigned to pretests and posttests containing a balanced mixture of items of each type written by each item writer. Dependent measures were (1) the pattern of pretest and Posttest item difficulties; (2) the variability of item difficulties across item writers; (3) the standard error of measurement of scales composed of items of each type; and (4) the concurrent validity of these scales. No significant differences between item writers were found in the analysis of item difficulties. Keyword nouns, which are relatively rare words in American textbooks but occur very frequently in a prose passage, were found to be unacceptable as question words. Algorithmic methods of foil writing were found to be feasible. (Author/CTM)