One question regarding the utility of adaptive testing is the effect of individualized item arrangements on examinee test scores. The purpose of this study was to analyze the item difficulty choices by examinees as a function of previous item performance. The examination was a 25-item test of basic algebra skills given to 36 students in an introductory statistics course at a large midwestern university during the fall 1985 semester. The test was administered via a microcomputer. Categorical data identifying the subjects' performance on current item, feedback condition, and choice of difficulty for next item were submitted to a log-linear analysis. A significant performance on current item by choice of difficulty for next item interaction was found. When examinees did not answer the current item correctly, they tended to request an easier item next; when they correctly answered the current item, they tended to request a harder item next. Preference for the difficulty of the next item appears to be a function of how well students believe they performed on the current item. Most adaptive testing item selection algorithms identify an easier item to be administered upon incorrect performance and a harder item subsequent to successful item performance. Results of this study suggest that this selection algorithm is congruent with examinee selected item difficulty. A table and a graph present study data. (Author/SLD)