Teacher-centered instruction has been widely adopted in college computer science classrooms and has some benefits in training computer science undergraduates. Meanwhile, student-centered contexts have been advocated to promote computer science education. How computer science learners respond to or prefer the two types of teacher authority, student-centered versus teacher-centered, has not yet been fully examined. Therefore, the purpose of the present study is to probe how preference for teacher-centered learning or student-centered learning may affect self-efficacy in learning computer science among undergraduate computer science majors. The participants were 804 college computer science majors across Taiwan. They anonymously responded to the "Teacher Authority Scale" (with four subscales: "Preference for autonomy," "Preference for participative management," "Preference for dependency," and "Preference for teacher control") and the "Self-efficacy in Learning Computer Science Scale." Structural equation modeling (SEM) was used to delineate two sets of variables. The results indicate that preference for autonomy in computer science learning positively predicts self-efficacy in learning computer science with the strongest coefficient. Computer science learner preference for teacher control is also a positive predictor. However, preference for participation in managing the computer class and preference for depending on the teacher did not play a significant role in the students' self-efficacy in learning computer science.