In this study we analysed what computational thinking skills pupils used when programming a small robot to solve a given problem, a simple programming task in which pupils had to select codes and put them in the correct place. Computational thinking is the process of (re)formulating a problem so that it becomes possible to use computer technology to solve it. Five computational thinking subskills were needed to solve the given problem: problem decomposition, abstraction, data ordering, testing and debugging. Forty-two pairs of pupils from grades 2 through 6 (ages 6-12) worked on the programming task. To capture their thoughts, they were asked to think aloud. The pairs were videotaped while working on the programming task. A rubric was developed to score the presence and quality of the use of the five subskills. Data analysis showed that just under half of the pairs succeeded in solving the task, mainly the pairs from the higher grades. For three of the five subskills (problem decomposition, abstraction, testing), the greatest improvement in approach occurred between placing code 1 and code 2. Successful pairs worked in a structured way.