In this work, we studied the influence of different programming approaches on the development of students' computational thinking (CT) skills, the programming experience and gender differences in CT development were also discussed. A total of 158 junior high school students and one teacher participated in the study over 5 months. The sample students were divided into four experimental groups in four single or combined programming approaches (i.e., plugged-in, unplugged, unplugged first, and plugged-in first) and one control group without programming. Data sources included the results of four CT tests, as well as interviews with the teacher and surveys with 24 representative participants. The results showed that the four programming approaches can effectively improve students' CT skills and can be retained after two months. Among them, the form of implementing unplugged activities before plugged-in can most effectively improve CT skills, and can better weaken the impact of previous programming experience. Finally, the qualitative analysis results provided insights into the process of programming and CT education. These findings will provide implications for the introduction of CT in junior high school, and help expand students' participation in computing.