Analysis and Evaluation of Liberal (Free-Choice) Multiple-Choice Tests

Abstract We present a new model of liberal (also known as free-choice) multiple-choice tests in which a candidate’s test score is represented as a function of two other variables: their inclination to eliminate perceived distracters; and the accuracy with which they are able to do this. We propose a number of theoretical relationships between these three variables that appear to support the notion that good students ought to be better judges of their level of knowledge — and hence better able to benefit through the expression of partial knowledge — than weak students, as previous empirical studies have found. However, the results of our own small empirical study were not in line with our expectations. This work has given us a better understanding of the test format, but we now doubt whether it does in fact provide the benefits that have previously been claimed.