Getting Lucky: How Guessing Threatens the Validity of Performance Classifications

Abstract

There is always a chance that examinees will answer multiple choice (MC) items correctly by guessing. Design choices in some modern exams have created situations where guessing at random through the full exam—rather than only for a subset of items where the examinee does not know the answer— can be an effective strategy to pass the exam. This paper describes two case studies to illustrate this problem, discusses test development decisions that can help address the situation, and provides recommendations to testing professionals to help identify when guessing at random can be an effective strategy to pass the exam.