Computational thinking (CT) skills help students both understand and utilize computing in various domains to cope with the challenges posed by the complex digital world. This study applied mind mapping to programming language instruction in a primary school, and explored the effects of mind mapping on students' CT skills when learning programming with Scratch as a programming tool. Two types of mind mapping, construct-by-self mind mapping (CBS-MM) and construct-on-scaffold mind mapping (COS-MM), were proposed to compare which type is more effective in terms of students' CT skills. Seventy-three public primary school fifth-grade students aged 10 to 11 participated in the study. The CBS group included 33 students learning with the CBS-MM instructional approach, while the COS group comprised 40 students learning with the COS-MM instructional approach. Data were obtained from nine topic Scratch programming assignments and pre- and post-CT skills surveys completed by the two groups. Paired sample t tests were used to analyze whether the two types of mind mapping approach could improve both groups' CT skills. One-way ANCOVA and one-way ANOVA were used to examine the difference in the two groups' CT skills. The results showed that both CBS-MM and COS-MM can improve students' CT skills; however, the COS group students could better improve their CT skills with the assistance of the COS-MM method compared with the CBS group students who learned with the CBS-MM method. The findings of this study can provide some enlightenment for relevant instructors to design instruction and implement programming teaching with different kinds of mind mapping support that facilitates students’ CT. •Computational thinking (CT) skills are important for students in the digital world.•Construct-by-self mind mapping (CBS-MM) can improve students' CT skills.•Construct-on-scaffold mind mapping (COS-MM) can improve students' CT skills.•COS-MM could better improve students' CT skills than those with CBS-MM.