Effects of Student Confidence and Item Difficulty on Test Score Gains Due to Answer Changing

Abstract

Two undergraduate and two graduate classes in statistics were given multiple-choice tests with subsequent evaluation of answer changes. The 95 students tested had an answer change rate of 6.6%. In evaluating the number of answer changes, no significant effect was found for ability, gender, nor the interaction between ability and gender. An analysis of gain scores due to answer changing showed a significant main effect for item difficulty, student change confidence, and their interaction. No significant effect on gain score was found for ability or any interaction with ability. Significant gains, even for changes based on low confidence, were interpreted as suggesting that previous cautions about answer changing are not warranted.