Random Guessing, Correction for Guessing, and Reliability of Multiple-Choice Test Scores

Guessing on a multiple-choice test item may be categorized as random (among all choices) or informed (some wrong choices eliminated). An expression is derived which permits determination of reliability change when random guessing is curtailed and informed guessing retained. Application of the derived expression to hypothetical and real data shows that small improvements in reliability will occur in this case except under highly implausible circumstances. Discussion follows on use of correction-for-guessing instructions to curtail random guessing. Possible reasons are given for frequently observed decreases in reliability when correction-for-guessing instructions are applied.