The effectiveness of a strategy for improving performance on multiple choice items for examinees with different levels of test anxiety was assessed. Undergraduate measurement students responded to the Mandler-Sarason Test Anxiety Scale and to an objective test covering course content. Results indicated that, for most examinees, generation of an answer before selecting a multiple choice response led to higher test performance, but for highly test anxious examinees, there was a negative effect on performance. Students with less test anxiety might have focused their attention more on the task at hand and thus used more effectively the thought processes involved in generating answers before approaching the multiple choice alternatives. This approach would be most beneficial to students who have real knowledge of the content being tested. Other variables of the test situation, such as impulsive answering, guessing, and other response sets might also be reduced. The experimental strategy resulted in smaller error of measurement and higher test reliability. (Author/GK)