Abstract: The purpose of this study was to examine the effects of control type (learner and program control) in computer-based instruction on middle-school students with varying levels of prior knowledge and preference for control. Ninety-nine sixth- and seventh-grade students participated in the study. The computer-based instructional program called Operations with Integers contained eight objectives that taught students how to add and subtract two one-digit integers. Students completed a 12-item pretest to measure prior knowledge and a 16-item Learner Preference Scale to indicate their preference for control. Students were blocked by preference-for-control scores and randomly assigned to either the learner-control or program-control version of instruction. Under program control, students viewed every example and practice-and-feedback opportunity. Under learner control, students could choose to bypass the remaining examples and practice-and-feedback opportunities for each objective after completing the first problem of each type. Analysis of achievement scores with a general linear model revealed a significant three-way interaction between prior knowledge, control type, and preference for control. The pattern of the interaction was that high-prior-knowledge students achieved relatively better on the posttest when their preference for control was matched with the type of control they received (high-preference-for-control received learner control and low-preference-for-control received program control), whereas low-prior-knowledge students achieved relatively better when their preference for control was mismatched with the type of control they received. High-preference-for-control subjects viewed significantly more optional elements than low-preference subjects and spent significantly more time in the program under both learner and program control. A 2x2x2 MANOVA on attitude scores also revealed a significant three-way interaction between control type, preference for control, and prior knowledge that reflected the same pattern found on the achievement data. High-prior-knowledge students had more positive attitudes when their preference for control was matched with the type of control they received, and low prior knowledge students had more positive scores when their preference for control was mismatched with the type of control they received. The findings indicate that high-prior-knowledge students have a tendency to achieve better and have more positive attitudes when they receive the type of control they prefer, whereas the opposite is true for low-prior-knowledge students.