Designed to develop a more practical model for adapting context to learner interests, this study used computer-based instruction to make alternative contexts for statistics problems--sports, business, education, or no-context--available for selection by individual learners. The subjects were 227 undergraduate students enrolled in required education courses at Memphis State University. The context adaptation strategy was evaluated as it was used both independently of and in combination with learner control (LC) of the practice examples. The research design involved the manipulation of standard (prescribed) contexts versus LC-context in combination with standard instructional support versus LC-instructional support. The study also investigated the nature of LC decisions in relation to the types of instructional options made available and learner characteristics. Instruments used in the study were the Nelson-Denney Reading Test, a unit pretest, a unit posttest, and a task attitude survey. Analyses of the data revealed little difference between the treatments in achievement; students who received minimum support found the lesson faster moving than those who received maximum support; 92% of the subjects had positive attitudes toward the desirability of selecting problem themes; and learners with higher prior achievement and reading ability tended to complete the task faster, score higher on the posttest, and view the task more favorably. It was concluded that, although the achievement results failed to support the hypothesized benefits for learning, a practical model for incorporating contextual adaptation in CBI had been developed and demonstrated. (37 references) (BBM)