The teacher-learner constellation is a special one in Human-Robot Interaction (HRI), as it can essentially improve intuitive interaction with robots. In a 2 (background: programmer vs. non-programmer) 3 (teacher: self vs. believed other vs. other) between participants experiment (n=48, counter-balanced in gender), participants kinesthetically taught a humanoid NAO robot a specific behavior, which the robot had to execute afterwards. Next, participants downloaded a taught behavior to the NAO and were told that the executed behavior either is (1) the one they previously taught (self), (2) the one someone else taught, but actually it was their own (believed other), or (3) the one someone else taught (other). We were interested in two main aspects: (1) whether programmers and non-programmers show differences in their teaching behavior and the perception of the teaching workload and (2) whether participants show a greater self-extension and trust into a robot they taught themselves over a robot they believed someone else taught. The study revealed that the teaching style independently of the background extends in the behavior execution time. Programmers showed a higher perceived workload than non-programmers. Differences in trust could not be found, but a self-extension effect was observed that people showed greater self-extension into a robot they taught themselves. Implications for Human-Robot Interaction are discussed.