The strategies examinees employ when making item difficulty level choices in self-adapted computerized testing were investigated. Subjects were 148 college students (88 females and 60 males) in an introductory statistics course. The primary instrument was a self-adapted computerized algebra test used to measure student readiness for the statistics course. Each examinee was administered 20 items from a pool of 93. Students rated their self-efficacy before the test and were administered measures of mathematics anxiety and test anxiety. Inspection of each student's data file provided an indicator of selection strategy. Examinees who chose a more difficult first test item expressed greater capability and higher confidence, reported less anxiety just prior to testing, and less anxiety about mathematics in general. When selecting additional items, examinees tended toward what was termed a sluggishly flexible strategy; they chose more difficult items after passing an item or string of items, and chose less difficult items after failing a single item or string of items. The most frequent choice was to remain at the same level. Results indicate that self-adaptive testing may be a viable alternative to computerized adaptive testing. Two figures and two tables contain data from the study. (SLD)