In this paper four classes of procedures for measuring the instructional sensitivity of reading comprehension test items are reviewed. True experimental designs are not recommended because some of the most important reading comprehension variables do not lend themselves to experimental manipulation. "Ex post facto" factorial designs are recommended because they allow the investigator to select and to study two or more variables and their interactions simultaneously. Correlational studies are recommended for the establishment of relationships between variables in the planning stages of an experimental study. Of the large family of instructional sensitivity indices reviewed, only one, the pre-to-post difference index, is recommended because of its demonstrated robustness and simplicity. (Author)