Searching for Better Scoring of Multiple-Choice Tests: Proper Treatment of Misinformation, Guessing and Partial Knowledge.

Brief explanations are presented of some of the different methods used to score multiple-choice tests; and some studies of partial information, guessing strategies, and test-taking behaviors are reviewed. Studies are grouped in three categories of effort to improve scoring: (1) those that require extra effort from the examinee to answer multiple-choice questions compared to the conventional mode of selecting the best or correct option among two or more choices; (2) those that manipulate test items or instructions; and (3) those that analyze original responses in special ways. The alternative scoring models of these studies have not provided a comprehensive picture of response behaviors because the manner in which they have been set up is fragmentary, and their assumptions are specific for the response mode for which the model was designed. A more generalized modeling approach is apparent in the finite state score theory of performance on multiple-choice tests proposed by M. A. Garcia-Perez (1987). Finite state score theory parsimoniously incorporates assumptions appropriate to the particular mode in which the test is administered, and it yields ability estimates on a single metric that remains the same regardless of the mode of test administration or assumptions concerning examinee behavior. The theory awaits empirical testing, but seems to be adaptable to any assumptions about item characteristics. Consequently, it appears to be a promising trend in measurement. A 31-item list of references is included. (SLD)