Pair programming (PP) is a useful approach to fostering computational thinking for young students. However, there are many factors impacting on the effectiveness of PP. The period of switching roles between the driver and the navigator is often ignored by researchers. Therefore, this study aimed to explore the impact of the switching period on PP. We conducted a PP experiment in four classes in the sixth grade in a primary school. The results indicated that (a) the semi-free switch was more effective for the learning achievement than the fixed periods, and the preference for adopting the fixed time interval to switch roles existed in previous studies seems to be a kind of prejudice; (b) students who switched roles every 5 minutes and semi-freely were more enjoyable than those who switched roles in every task and in every class session. Moreover, the period of switching roles in every class session decreased students' enjoyment after PP; (c) the frequency of switching roles decreased significantly, but the negotiation between the driver and the navigator became more active with time going in the semi-free class. Implications for teaching are also discussed.