It has been found that when examinees are allowed to choose a subset of constructed response (CR) items to answer on a test, they tend to choose differently and often perform lower on more popularly chosen items. The purpose of this study was to examine this finding. Using an experiment that incorporated a revised Advanced Placement (AP) Chemistry examination and a well-designed questionnaire, this study revealed a series of psychological factors that consistently influence examinee's choices of CR items. The instrument package was administered to 680 students in Hawaii at 4 levels of college and high school chemistry study. The national choice pattern of the five essays was well replicated by the Hawaiian students, indicating the existence of a consistent influence on examinees' choices. This influence was attributed to the commonality of chemistry curricula and textbooks. Students' perception of essay difficulty did accurately predict and account for the popularity of essay choices, and their perception of content dimensionality of essays coincided with their choices. The findings from this study offer a number of suggestions regarding CR item pretesting and test construction and a wide range of application possibilities for performance-oriented examinations. An appendix contains the survey and test kit. (Contains 3 tables, 4 figures, and 15 references.) (Author/SLD)